PROCESSING OF THE MULTIGROUP CROSS-SECTIONS FOR

MCNP CALCULATIONS

JAKUB LÜLEY1,2, BRANISLAV VRBAN1, ŠTEFAN ČERBA1, FILIP OSUSKÝ1,

VLADIMÍR NEČAS1
1
Slovak University of Technology in Bratislava, Faculty of Electrical Engineering and
Information Technology, Institute of Nuclear and Physical Engineering, Ilkovičova 3, 812 19
Bratislava, Slovakia
E-mail: [email protected], [email protected], , [email protected],
[email protected]
2
B&J NUCLEAR Ltd., Alžbetin Dvor 145, 900 42 Miloslavov, Slovakia
E-mail: [email protected]
Abstract: Stochastic Monte Carlo (MC) neutron transport codes are widely used in various
reactor physics applications, traditionally related to criticality safety analyses, radiation
shielding and validation of deterministic transport codes. The main advantage of Monte
Carlo codes lies in their ability to model complex and detail geometries without the need of
simplifications. Currently, one of the most accurate and developed stochastic MC code for
particle transport simulation is MCNP. To achieve the best real world approximations,
continuous-energy (CE) cross-section (XS) libraries are often used. These CE libraries
consider the rapid changes of XS in the resonance energy range; however, computing-
intensive simulations must be performed to utilize this feature. To broaden our computation
abilities for industrial application and partially to allow the comparison with deterministic
codes, the CE cross section library of the MCNP code is replaced by the multigroup (MG)
cross-section data. This paper is devoted to the cross-section processing scheme involving
modified versions of TRANSX and CRSRD codes. Following this approach, the same data
may be used in deterministic and stochastic codes. Moreover, using formerly developed and
upgraded cross-section processing scheme, new MG libraries may be tailored to the user
specific applications. For demonstration of the proposed cross-section processing scheme,
the VVER-440 benchmark devoted to fuel assembly and pip-by-pin power distribution was
selected. The obtained results are compared with continues energy MCNP calculation and
multigroup KENO-VI calculation.
Keyword: MCNP, Multigroup calculation, VVER, criticality

1. INTRODUCTION
A lot of effort has been spent on the development of techniques to effectively compute
sensitivity coefficients and cross-section induced uncertainties (S/U) by Monte-Carlo codes
and continues-energy libraries. MCNP6 and SCALE6.2 have currently implemented
approaches to calculate the adjoint weighted tallies, which allow both codes to carry out S/U
analyses related to criticality safety calculations using CE libraries. Although, it is no
necessary for CE libraries to calculate flux moments from mesh flux, implicit sensitivity
coefficients or adjoint flux and to perform cell treatment, S/U calculation can be still memory
and time expensive and special attention must be given to the preparation of the covariance
matrix. [1,2,3] Alternatively, the multigroup option is still an effective method for other
applications and in case of cross-section adjustment, it can also extend the applicability of S/U
calculation. The main advantage of multigroup calculation is not only in the reduction of the
calculation time but also offers to an engineer utilization of the adjoint calculations in
problems where the forward transport calculation is not efficient. [4]
Effective use of the multigroup approach is depending on the appropriate and target
tailored cross-section library. To achieve a required accuracy of the multigroup MC
calculation, the problem specific cross-sections have to be available. Currently, there are four
VVER-440 reactor units in operation and two other units are under construction in Slovakia,
thus there is a clear need of improvement in our calculation abilities. The VVER-440 reactors
belong to PWR family, where thermal scattering treatment has to be taken into account during
cross-section data processing. In many industrial applications, a validation of deterministic
codes requires that the MC code utilize the same data as a deterministic code [5]. Optimized
cross-section library applicable to both types of calculation is therefore the basic requirement
for the current neutronic analyses and allows us to implement various methods to the
multigroup constants processing scheme.
2. THE MULTIGROUP CROSS-SECTION PROCESSING
2. 1. Cross-section library
Within this analysis the SBJ_V2019T multigroup cross-section library, which is
updated version of SBJ_V2018T library presented in [5], was utilized as source data file. The
library is based on the ENDF/B-VII.1 evaluated data [6], includes thermal scattering data for
hydrogen in H2O based on IKE S(α,β) [7, 8] and CAB [9] models; and is stored in MATXS
format. The cross-section library was prepared in 238 group structure, the same as in
SCALE6 libraries [10], where the core averaged neutron spectrum of the VVER-440 reactor
was used as a weighting function. Utilization of the SBJ_V2019T library is a part of long
term complex validation process of developed cross-section processing scheme.
2. 2. Computational scheme
The fundamental part of the problem oriented MG constants processing scheme is a
universal code which is able to prepare MG constant in various formats. In our previous
analyses, the TRANSX code [11] was standardly utilized. The TRANSX code is able to work
with cross-section library in general MATXS format and region-wise flux files in CCCC
format. It can process cross-section data with an appropriate cell treatment; infinite
homogeneous mixture incorporating self-shielding effects by Bondarenko Method or lump
materials incorporating Dancoff correction for several geometrical configurations (slab or
cylinders in triangular or square lattice). Some modifications were made in the TRANSX
source code to enhance its versatility, like considering cladding within Dancoff correction
calculation, enabling utilization of dummy materials, extraction of scattering matrixes for
individual reactions (elastic, inelastic and n2n) and others. The TRANSX code prepares
problem oriented MG constants (micro or macroscopic) standardly in ISOTXS format, which
can be directly used by deterministic codes PARTISN [12] or DIF3D [13]. To process MG
constants for MCNP calculation the CRSRD [4] modules were implemented to the TRANSX
code. The CRSRD code was designed to translate deterministic MG cross-section to the
format suitable for MCNP.
 Figure 1. Computational scheme
The computational scheme applied in this analysis is shown in Fig. 1. The left side is
devoted to the processing of the SBJ_V2019T library which is describe in [5]. The right side
is aimed to the transport calculations and is described in following sections. The mid part is
schematically describing the data flow and the utilized codes. In the first round the
SBJ_V2019T library is directly used to prepare two sets of MG cross-sections. The first set
was prepared in the way that all materials were processed as an infinitely diluted, in 238
group structure and in MCNP MG format, later denoted as 238gH. The second set was
prepared in the way that all structural materials were processed as infinitely diluted, in 238
group structure and in MCNP MG format. Fuel, cladding and coolant were processed through
the cell treatment considering cylindrical shape of lump in triangular lattice and in 238 group
structure in MCNP MG format, later denoted as 238g and in ISOTXS format for spectral pin
calculation. In the second round the region averaged flux spectra calculated for each fuel
enrichment was used to collapse MG cross-sections. The 44 group structure, same as used in
SCALE6 libraries, was used due to its compatibility with 238 energy group structure. The
third set is later denoted as 44g. All three cross-section sets comprise of isotope-wise
constants (microscopic cross-section) and homogenized material-wise constants (macroscopic
cross-sections). Calculations, which utilized homogenized constants are later denoted with
abbreviation HM as 238gHHM, 238gHM and 44gHM respectively to the individual sets.
All the TRANSX calculations were performed with the same set of input parameters.
Legendre polynomial expansion for scattering matrixes was set to 3. Consistent-P
approximation was used for transport correction and 95 thermal groups were defined in the
238 group calculation. In the 44 group calculation, for transport correction, the Bell-Hansen-
Sandmeier approximation was used and 23 thermal groups were defined. In case of CRSRD
translation part, the absorption cross-sections were kept with negative values, if occurred, and
scattering matrixes were processed with the factor 2ℓ+1 to 32 equi-probable bins by the
Maximum Entropy approach. Since this process was managed by the TRANSX code with
implemented modules from CRSRD, the data flow and the formats from the MATXS through
TRANSX and CRSRD procedures to MCNP MG cross-section file were controlled. More
information with the data flow scheme is presented in [14].
All MCNP MG calculations were validated by the MCNP calculation in CE mode and
KENO-VI [10] MG calculation using 238 group library. In case of MCNP CE calculation the
cross-section data processing is quite straightforward. Regarding to KENO-VI calculation all
parameters related to the cross-section data processing were selected similarly as much as it
was possible, like 238 group library, cell treatment, Legendre expansion, material definition,
etc.
2. 3. VVER-440 benchmark overview
The VVER-440 benchmark devoted to fuel assembly and pip-by-pin power distribution
was selected to demonstrate the proposed cross-section data processing capabilities.
Benchmark definition is based on zero power state with uniform temperature 543.15 K. The
geometry was simplified to two-dimensional definition with 30-degree symmetry. All internal
parts like core barrel, core basket and core rim called “vygorodka” were included. The outer
boundary of the benchmark is defined by the outer edge of reactor pressure vessel with
vacuum boundary condition. The geometry definition is presented in Fig. 2, based on the
SCALE6 model. The core is composed of three types of fuel assemblies with enrichment 1.6,
2.4 and average 4.25 % and two fully inserted Emergency Reactor Control Assembly (ERC)
of 6th working group. The fuel assembly with average enrichment 4.25 % consists from fuel
pins with four different enrichments where one type of pin contains combination of the UO2
and gadolinium absorber. All fuel assemblies are considered as a fresh fuel. [15]
 Figure 2. VVER-440 benchmark geometry model [16]

3. RESULTS AND DISCUSSION

The first type of calculation was carried out at the level of fuel pin to obtain neutron
flux spectra for cross-section collapsing. These calculations were performed by the PARTISN
code in simplified 1D geometry where the hexagonal lattice cell was represented by the radius
of equivalent cylinder. Due to direct utilization of flux files (RZFLUX shown in Fig. 1) and
capabilities of the TRANSX code, helium volumes were omitted in the geometry model. The
same geometry definition, except hexagonal to cylindrical approximation, was used also in
the comparative MCNP and KENO-VI calculations. MCNP as well as KENO-VI was run
with 10 mil. histories in 1000 generation and 30 generations were skipped.
The main part of the research is focused on the validation of proposed cross-section
processing scheme utilizing in-house prepared cross-section data and software application on
benchmark calculation, where not only integral parameter keff is evaluated, but also local
parameters like relative fuel assembly (FA) power distribution and relative fuel pin (FP)
power distribution are available. Within this paper only results of keff and FA power
distribution are presented and discussed. Geometrical and material model of the benchmark
core was prepared for MCNP and KENO-VI codes as it is described in the previous part and
shown in Fig. 2. All MCNP calculations were performed by the MCNP5 1.6 version [17] with
750 mil. neutron histories within 5 000 generations. The KENO-VI calculation was performed
using the version distributed within the SCALE6.1.3 system with 100 mil. neutron histories
during 500 generations.
Due to the relatively small difference between the results of IKE and CAB thermal
scattering models (in maximum it was 63 pcm) only the results of CAB model are presented
in the next chapter. All comments and conclusions presented in this paper are relevant and
applicable for the results of IKE model.
3. 1. Results
The results of the FP spectral calculations are presented in Table 1. Six different
enrichments represent three types of FA where the last four belong to the FA with average
enrichment 4.25 %. The results of spectral calculation carried out by PARTISN code are
supplemented by the MCNP CE and KENO-VI MG calculation due to only partial validation
of the SBJ_V2019T library and the part of cross-section processing dedicated only to
deterministic codes. The first value presented for each code is the infinite multiplication factor
kinf, instead of keff, due to reflective boundary conditions on geometric model edges. The
second value in case of MC codes stands for the standard deviation of kinf.
 Table 1. Comparison of fuel pin (FP) calculation

FP enrichment PARTISN MCNP CE KENO-VI MG

% of U235 kinf kinf σ kinf σ
1.6 1.06290 1.07071 0.00018 1.06392 0.00015
2.4 1.18125 1.18943 0.00019 1.18228 0.00019
3.6 1.27898 1.28595 0.00020 1.27894 0.00020
4.0 1.30090 1.30790 0.00020 1.30036 0.00018
4.0+3.35Gd2O3 0.39722 0.39874 0.00008 0.39647 0.00009
4.4 1.31954 1.32602 0.00021 1.31864 0.00019

Values of keff from the MC VVER-440 benchmark calculation are presented in Table 2.
The first three values can be considered as a reference due to their independence on the
demonstrated cross-section scheme processing and the previous validation. The next six
values represent the current capabilities of the cross-section processing scheme and
possibilities of future use.
Table 2. Results of keff of the VVER-440 benchmark

Computational case keff σ

MCNP ref [15] 1.06827 0.00005
KENO-VI 238g [16] 1.06343 0.00006
MCNP CE 1.06478 0.00002
MCNP MG 238gHM 1.05447 0.00006
MCNP MG 44gHM 1.05498 0.00008
MCNP MG 238g 1.05161 0.00002
MCNP MG 44g 1.05015 0.00002
MCNP MG 238gH 1.08597 0.00002
MCNP MG 238g HHM 1.08831 0.00006

The relative FA power distribution calculated by the KENO-VI and MCNP in CE and
MG mode with their relative change from the reference MCNP calculation are presented in
Fig. 3, Fig. 4 and Fig.5 respectively. MCNP MG case is represented by the calculation with
44g collapsed data, since it is the most perspective one for application in future analyses.

Figure 3. Relative FA power distribution of KENO-VI (left) calculation and relative change
in % from MCNP reference values (right)
 Figure 4. Relative FA power distribution of MCNP CE (left) calculation and relative change
in % from MCNP reference values (right)

Figure 5. Relative FA power distribution of MCNP MG 44g (left) calculation and relative
change in % from MCNP reference values (right)

3. 2. Discussion
Comparison of the fuel pin calculation demonstrates the validity of the SBJ_V2019T
library for thermal applications and the fact that the whole functionality of TRANSX code has
been maintained after the implementation of CRSRD modules. Special attention should be
given to the comparison of MG calculations (PARTISN vs. KENO-VI, see Table 1) where the
relative difference in almost all cases is less than 100 pcm. Only for calculation of the FP
consisting Gd the relative difference was higher, but 479 pcm is still quite reasonable due to
comparison of very low keff. Relative difference between PARTISN and MCNP CE
calculation is also acceptable because relative difference in average is just 570 pcm and
PARTISN calculations systematically underestimated the MCNP CE calculations.
Within the criticality calculation of the VVER-440 benchmark, the first inconsistency in
the results presented in Table 2 was obtained at the level of reference calculations. The
relative difference between the reference MCNP and KENO-VI or MCNP CE is in both cases
more than 300 pcm. The obtained difference has been caused by the different nuclear
evaluated data used during libraries processing. Reference MCNP calculation was carried out
with the library processed from ENDF/B-VI data, while the KENO-VI library is based on
ENDF/B-VII.0 and MCNP CE library is based on ENDF/B-VII.1 data. Therefore, it is more
suitable to compare calculated values of MCNP MG with KENO-VI or MCNP CE values, at
least at the level of keff.
The relative difference between MCNP MG and MCNP CE calculations vary from 872
pcm to 2031 pcm in absolute values. The closest keff was obtained for cases where the
homogenized macro-constants were used with appropriate cell treatment (44gHM and
238gHM). Relatively comparable values were obtained from calculation with isotope-wise
constants (44g and 238g). Significant difference is in the calculated standard deviation. While
the computational time of cases with homogenized macro-constants is around 10 % faster
than in cases with micro-constants, the obtained standard deviation is more than three times
higher. Therefore, all promising benefits from a utilization of homogenized macro-constants
is vanishing in this stage of development. Generally, utilization of MG constants shortened
the computational time about 30 %. The worst results were obtained for the cases with
materials which were all treated as infinitely diluted.
In case of calculation of relative FA power distribution, inconsistent results were
obtained. While the relative change between KENO-VI or MCNP CE values and reference
MCNP values is in maximum 3.1 % for KENO-VI and 6.8 % for MCNP CE, the relative FA
power distribution of MCNP MG calculations compared with reference MCNP is
unacceptable. The relative change varies from -42 % to 17 % which is suggesting that the MG
constants are not properly prepared. Revision of all input files and source code of the
TRANSX/CRSRD code did not reveal any formal or technical problem. But collected
information are pointing to different interpretation of the absorption cross-section in
TRANSX and MCNP.
4. CONCLUSION
This paper demonstrates the first result of cross-section processing scheme aimed to
preparation of the problem oriented MG cross-section constants for MCNP calculation. The
first stage of research activities involved development of the new code able to prepare MG
constants for deterministic as well as MC codes, which is almost done. Although, the first
result of benchmark calculation with complex geometry identified relevant problems with the
cross-section interpretation, quite good experience with TRANSX code itself and experience
with cross-section processing for deterministic codes give us opportunity to solve this
problem in near future. The achieved reduction of computational time about 30 % is
promising for next applications, but significant difference between 238g and 44g was not
observed. The following analysis will be therefore focused on this phenomenon and optimal
coarse group structure will be sought. The performance of the SBJ_V2019T library fulfilled
all requirements and its applicability at least within deterministic calculations was clearly
demonstrated on pin spectral calculations. Next step will be the final validation on the
complex set of criticality safety benchmarks
5. ACKNOWLEDGEMENT
This study has been partially financially supported by the Slovak Research
Development Agency No. APVV-16-0288 and by the Scientific Grant Agency of the Ministry
of Education of Slovak Republic No. VEGA 1/0863/17. We would like to also acknowledge
the ongoing support of B&J NUCLEAR ltd. regarding to the code development and sharing
necessary information.
6. REFERENCES
1. B. C. Kriedrowski, F. B. Brown, “Applications of Adjoint-Based Techniques in
Continuous-Energy Monte Carlo Criticality Calculations”. Joint International
Conference on Supercomputing in Nuclear Applications and Monte Carlo 2013, Paris,
France, 2013
2. C.J. Werner(editor), "MCNP User‟s Manual - Code Version 6.2", Los Alamos
National Laboratory, report LA-UR-17-29981, 2017
3. Oak Ridge National Laboratory, “SCALE Code System: Version 6.2.2.”, ORNL/TM-
2005/39, p. 2747, 2017
4. J. C. Wagner, E. L. Redmont II, S. P. Palmtag, J. S. Hendricks, „MCNP:
Multigroup/Adjoin Capabilities, “ Los Alamos National Laboratory, 1994
5. Š. Čerba, B. Vrban, J. Lüley, F. Osuský, V. Nečas, "Development of Multi-Group XS
Library for VVER-440", Proc Int. Conf. Nuclear Energy for New Europe, Portorož,
Slovenia, 2018
6. M. B. Chadwick, M. Herman, P. Obložinský, et al., “ENDF/B-VII.1 Nuclear Data for
Science and Technology: Cross Sections, Covariances, Fission Product Yields and
Decay Data” in Nuclear Data Sheets, vol. (112), pp. 2887-2996, 2011.
7. Š. Čerba, J. I. M. Damian, J. Lüley, B. Vrban, V. Nečas, "Comparison of thermal
scattering processing options for S(aplha,beta) cards in MCNP", Annals of Nuclear
Energy, vol. 55, pp. 18-22, 2013
8. J. Keinert, M. Mattes, E. Sartori, "JEF-1 scattering law data", JEF Report 2/JEF/DOC
41.2. Tec. Rep., IKE Stuttgart, 1984
9. J. I. Márquez Damián, J. R. Granada, D. C. Malaspina, “CAB models for water: A
new evaluation of the thermal neutron scattering laws for light and heavy water in
ENDF-6 format”, Annals of Nuclear Energy, vol. 65, pp. 280-289, 2014
10. SCALE: A Comprehensive Modeling and Simulation Suite for Nuclear Safety
Analysis and Design, ORNL/TM-2005/39, Version 6.1.3, Oak Ridge National
Laboratory, Oak Ridge, Tennessee, 2011
11. R.E. MacFarlane, “TRANSX 2.15: A Code for Interfacing MATXS Cross-Section
Libraries to Nuclear Transport Codes for Fusion Systems Analysis”, Los Alamos
National Laboratory, Los Alamos, 1992.
12. R. E. Alcouffe, R. S. Baker, J. A. Dahl, S.A. Turner, and Robert Ward, “PARTISN: A
Time-Dependent, Parallel Neutral Particle Transport Code System,” LA-UR-08-
07258, 2008
13. K. L. Derstine: DIF3D 10.0, Argonne National Laboratory, Argonne, IL, 2011
14. B. Vrban, Š. Čerba, J. Lüley, F. Osuský, V. Nečas, J. Haščík, ", Proc Int. Conf.
Nuclear Energy for New Europe, Portorož, Slovenia, 2018
15. V. Krýsl, P. Mikoláš, D. Sprinzl and J. Švarný “„Full-Core‟ VVER-440 benchmark
extension”, Proc. of the 24th Symposium of AER, Sochi, Russia, 2014
16. J. Lüley, B. Vrban, Š. Čerba, F. Osuský, V. Nečas, J. Haščík, “Assembly
homogenization technique for VVER-440 reactor multi-group calculations” ", Proc
Int. Conf. Nuclear Energy for New Europe, Portorož, Slovenia, 2018
17. LANL, „MCNP - A General N - Particle Transport Code“, Los Alamos National
Laboratoy, Los Alamos, 2003