Multigroup Cross-Section Libraries for MCNP

Branislav Vrban
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
[email protected]

Štefan Čerba, Jakub Lüley, Filip Osuský, Vladimír Nečas, Ján Haščík
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
[email protected], [email protected], [email protected],
[email protected], [email protected]

ABSTRACT

Stochastic Monte Carlo 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 to accept
simplifications. To achieve the best real world approximations, continuous cross section
libraries (CE) are often used. These CE libraries take into account the rapid changes of XS in
the resonance energy range; however, computing-intensive simulations must be performed to
utilise this feature. Currently, one of the most accurate and developed stochastic MC code for
particle transport simulation is MCNP, which is widely used at Institute of Nuclear and
Physical Engineering FEI STU in Bratislava. To broaden our computation abilities and to
allow the comparison with deterministic codes, the CE cross section library of the MCNP
code is replaced by the multigroup cross section data (MG), generated by the modified
versions of TRANSX and CRSRD codes. Following this approach, the same data may be
used in deterministic and stochastic codes. Moreover, by the use of formerly developed and
upgraded cross section processing scheme, new MG libraries may be tailored to the user
specific applications. The paper represents the first benchmarking of the newly developed
multigroup cross section processing for the MCNP code, the code abilities and the identified
discrepancies. Finally, some discussions on further needed development are also included.

1 INTRODUCTION

The principal role of Monte Carlo (MC) codes is to simulate the real highly-detailed
geometry reactor systems based on known probabilities of occurrence with random sampling
of the Probability Functions. Among MC codes, MCNP [1] is a worldwide general purpose
Monte Carlo code which has very strong ability to treat complex nuclear interaction
processes. However, some complicated models can be time consuming due to using the
continuous cross section (CE) libraries. The multigroup and homogenised cross section data

227.1
 227.2

(MG) has been already successfully used for analysis in reactor physics, thus their utilisation
to speed-up the simulation is a natural choice. A standard MG set is available for MCNP
coupled neutron-photon transport. However, MG sets are application dependent; therefore
there is a need for users to generate their own specific libraries. Moreover, besides the speed-
up efforts the comparisons with deterministic codes and cross-section sensitivity and
uncertainty studies require that the Monte Carlo code utilise the same data as deterministic
codes. To improve our computation abilities and to allow the comparison with deterministic
codes, the CE cross-section library of the MCNP code is replaced by the MG section data
generated by the modified versions of TRANSX [2] and CRSRD [3] codes. To demonstrate
the capabilities of the newly developed code a few high-quality integral experiments from the
International Handbook of Evaluated Criticality Safety Benchmark Experiments (ICSBEP
Handbook) [4] and International Handbook of Evaluated Reactor Physics Benchmark
Experiments (IRPhE) [5] are evaluated.

2 APPLIED TECHNIQUES AND METHODS

2.1 Cross-section processing

To perform a neutron transport calculation with multiple nuclides in a given set of

materials, the first processing step that must be taken is to represent all of the evaluated cross-
section data (XS) (generally stored in Evaluated Nuclear Data Formats – ENDF-6) on a grid
common to all of the nuclides in the problem [6]. In terms of multigroup calculation, the
energy range is divided to the finite set of energy groups, over which the cross-sections are
assumed constant. One of the most comprehensive and popular code system used to represent
ENDF/B data on a common energy grid is NJOY originally being developed by the Los
Alamos National Laboratory (LANL). After converting all nuclides from parametric
representation used in ENDF/B to a common energy grid, NJOY generates the actual
resonance range XSs for a two-dimensional matrix of temperature and background scatter
cross-sections. These data are then stored to a MATXS (for material XS library) file. In our
XS processing scheme (shown in Figure 1 by green colour), the code TRANSX [2] is used for
a given problem to interpolate MATXS data in the resonance grid to get a problem-dependent
self-shielded XSs for each material.

Figure 1: Overview of the cross-section processing paths

These data are then automatically provided in the DTF library directly to the CRSRD,
which processes deterministic MG XS libraries into a format suitable to MCNP. The details

Proceedings of the International Conference Nuclear Energy for New Europe, Portorož, Slovenia, September 10 – 13, 2018
 227.3

of XS processing path for MCNP CE (bounded by the dashed blue line) calculations can be
found in our previous works [7], [8], and the procedure related to production of MG XS
libraries for deterministic codes (dashed red line) is thoroughly described in [9], [10].

2.2 Criticality benchmark problems

Although MCNP has been thoroughly benchmarked in the continuous energy mode at
our Institute [8], it has not been adequately validated in the multigroup mode. Some results
based on subset of problems specified in [11] are available in [3]; however these problems are
geometrically quite simple. Our first approach to demonstrate MCNP MG capabilities
comprises the criticality calculations of a few high-quality integral experiments from the
ICSBEP Handbook and IRPhE database. The first selection of experiments was based on the
approach shown in [12], where the selected experiments cover a wide range of fast neutron
energy spectra, and include critical masses, spectral indices and void reactivity coefficients.
The list of the selected experiments with their abbreviations used in this paper is shown in
Table 1.
Table 1: The list of integral experiments with experimental quantities used in analysis
Integral experiments Handbook name 𝑘𝑒𝑓𝑓 𝜎
FLATTOP Pu configuration (FLATTOP) PU-MET-FAST-006 1.000000 1.60E-03
Jezebel 239Pu configuration (Jez239) PU-MET-FAST-001 1.000000 2.00E-03
Jezebel 240Pu configuration (Jez240) PU-MET-FAST-002 1.000000 2.00E-03
ZPPR9 ZPPR-LMFR-EXP-002 1.001060 1.17E-03
JOYO JOYO-LMFR-RESR-001 1.001100 1.80E-03
ZPR6-7 standard configuration (ZPR6-7st) ZPR-LMFR-EXP-001 1.000510 8.70E-04
ZPR6-7 high 240Pu content (ZPR6-7hp) ZPR-LMFR-EXP-002 1.000800 9.00E-04

To address the needs of MG XS resonance self-shielding processing very detailed

computational models of the last four experiments shown in Table 1 are used in parallel with
their simplified versions (denoted by superscript S in results section). The material and
geometry definitions can be found on the WPEC SG33 web page [12]. The different level of
heterogeneities used in the integral experiments is demonstrated in Figure 2.

a) FLATTOP [4] b) JOYO MCNP model c) ZPPR [4]

Figure 2: The different level of heterogeneities in integral experiments

Proceedings of the International Conference Nuclear Energy for New Europe, Portorož, Slovenia, September 10 – 13, 2018
 227.4

2.3 Calculation details

With the aim to validate ability to produce MG data for MCNP, the MATXS data were
directly taken from ZZ-KAFAX-E70 [13] nuclear data libraries. This approach excludes
possible errors in NJOY processing and gives more attention to the next processing steps. The
ZZ-KAFAX-E70 library uses 150 group structure for neutron data and was prepared based on
the ENDF/B-VII.0 [14] evaluated data file weighted by KALIMER-150 (Korean Advance
LIquid MEtal Reactor) equilibrium core flux.
The XS multigroup approximation necessarily yields to discrepancies related to the used
coarse energy structure and real resonances of used isotopes. The presence of resonance
material in a mixture can cause large neutron flux perturbations and may affect the related
reaction rates. This energy (resonance) self-shielding effect can also significantly affect
calculation results of multiplication properties. Therefore, the principally two approaches
investigating the influence of the background XS [15] are studied in this paper. In the first
approach, the MG XS data for MCNP were prepared isotope-wise, where no influence of
other materials in the material was taken into account. Thus, the final library contained all of
the needed isotopes processed to an appropriate temperature (Doppler broadening) regardless
of its location in specific material. In this approach no data about real material background XS
is available for modified version of TRANSX 2.15 code processing. The second approach
reflects the real composition (and thus background XS for specific isotope and resonance) of
each benchmark material. Due to complicated structure and high number of materials used in
the selected integral experiments, a new in-house tool able to automatically extract material
data and prepare appropriate TRANSX inputs was developed. Accordingly, the final XS
library prepared for specific benchmark comprises more sets of data for a particular isotope,
where each set is related to one material in MCNP input file.
Another important input to XS processing in TRANSX code from the view of
heterogeneities is the value of escape XS. The escape XS is defined by the inverse of the
average chord length. It can be seen [16] that the sample mean chord length is a function of
sample surface area and volume, and does not depend on the shape of the sample. However,
this is exactly true just for convex samples and isotropic flux distribution. In our approach and
taking into account fast spectrum of integral experiments and their diverse geometries, the
mean chord length was set to artificial value of 10 000 cm. By this approach we set a low
escape probability of neutrons, so the neutron flux depression in the material with resonances
becomes higher.
To run CRSRD calculation, several parameters have to be set by the user. In our case
the negative absorption cross-sections were set to zero because the analog capture has been
used in MCNP calculation. The Maximum Entropy approach [17] was used for angular
scattering conversion from the Legendre expansion to 32 equi-probable bins. The calculation
was executed with the Legendre order of 3. All final MC calculations were performed by
MCNP5 1.60 in multigroup mode.

3 RESULTS

The Figure 3 compares multigroup and continuous energy MCNP results of effective
multiplication factors for the subset of above defined integral experiments. It should be noted
that the one sigma uncertainty in all cases were kept bellow 𝜎𝑘𝑒𝑓𝑓 < 2.5 × 10−4 . MG libraries
are target dependent, therefore the comparison to experimental values is not explicitly

Proceedings of the International Conference Nuclear Energy for New Europe, Portorož, Slovenia, September 10 – 13, 2018
 227.5

provided. However, if a reader is interested, the experimental values of effective

multiplication coefficients can be found in Table 1. The NOBGXS (No Back Ground Cross
Section) abbreviation used in the figure stands for the first approach defined in calculation
details paragraph. The BGXS denotes to the second approach where other isotopes in the
mixture were taken into account in XS processing.

Figure 3: Comparison of the keff values from the MG and CE MCNP calculations
The results shown in Figure 3 verify that MG MCNP compares well (within 1 %) with
CE calculations for the geometrically simple experiments (FLATTOP, JEZ239 and JEZ240).
The discrepancies in other experiments can be attributed to the constant escape probability
approach and mean chord length used as an input to TRANSX calculations. It is evident, that
the second approach taking into account the real background scatter cross-sections of each
material made MG results much closer to CE results, especially in case of simplified models
(from 5 % to below 1 %).

4 CONCLUSION

The MCNP MG option is valuable for comparisons with deterministic codes, running
adjoint problems, speed-up of calculations, sensitivity and uncertainty studies and allows the
simulation of charged particles. The XS processing scheme shown in paper allows the full
control of the nuclear data to fulfil desired users’ needs. The results demonstrated are
promising, however much more integral experiments need to be added to the benchmarking
subset. As it was shown in the calculation details section, several parameters should be set up
by user; therefore to assess their impact, a parametric study will be performed in near future.
It is clear, that the S(α, β) thermal scattering treatment is inappropriate in the MG mode, and
should be fully treated in the MG library. The work on TRANSX modification allowing the
use of this specific data is already ongoing at our institute and the first results seem to be quite
promising. After the correct implementation of such a calculation option, highly moderated
experiments will be added to the MCNP MG benchmarking subset. Possibly more options
allowing the calculation of effective mean chord length also assuming anisotropic neutron
flux will be added to enhance our computational abilities.

Proceedings of the International Conference Nuclear Energy for New Europe, Portorož, Slovenia, September 10 – 13, 2018
 227.6

ACKNOWLEDGMENTS

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.

REFERENCES

[1] LANL, "MCNP - A General N - Particle Transport Code," Los Alamos National
Laboratory, Los Alamos, 2003.

[2] R. E. MacFarlane a D. C. George, „RSICC Peripheral Shielding Routine Collection

TRANSX 2.15,“ ORNL, 1995.

[3] J. C. Wagner, E. L. Redmont II, S. P. Palmtag et al, „MCNP: Multigroup/Adjoin

Capabilities,“ Los Alamos National Laboratory, 1994.

[4] Nuclear Energy Agency, „International Handbook of Evaluated Criticality Safety

Benchmark Experiments,“ OECD, 2010.

[5] OECD NEA, „International Handbook of Evaluated Reactor Physics Benchmark

Experiments,“ OECD Nuclear Energy Agency, 2012.

[6] B. Zohuri, Neutronic Analysis For Nuclear Reactor Systems, Albuquerque: Springer,
2017.

[7] Š. Čerba, J. I. M. Damian, J. Lűley et al, „Comparison of thermal scattering processing

options for S(alpha,beta) cards in MCNP,“ Annals of Nuclear Energy, zv. 55, pp. 18-22,
2013.

[8] J. Lüley, B. Vrban, G. Farkaš et al, „Determination of criticality safety MCNP5

calculation bias by using different libraries of cross section,“ Progress in Nuclear
Energy, zv. 59, pp. 96-99, 2012.

[9] Š. Čerba, B. Vrban, J. Lűley et al, „Multigroup cross section library for GFR2400,“ rev.
Proceedings of ND 2016: International Conference on Nuclear Data for Science and
Technology, 2017.

[10] B. Vrban, J. Lüley, Š. Čerba et al, „Cross section adjustment for fast reactor design:
ATCROSS code and its application,“ Annals of Nuclear Energy, zv. 110, pp. 928-940,
2017.

[11] D. J. Whalen, A. D. Cardon, J. L. Uhle et al, „MCNP: Neutron Benchmark Problems,“

Los Alamos National Laboratory.

[12] NEA, OECD, „Methods and Issues for the Combined Use of Integral Experiments and
Covariance Data,“ OECD, 2013.

Proceedings of the International Conference Nuclear Energy for New Europe, Portorož, Slovenia, September 10 – 13, 2018
 227.7

[13] D. H. Kim , C. S. Gil and Y. O. Lee, „ZZ KAFAX-E70, 150 and 12 Groups Cross
Section Library in MATXS Format based on ENDF/B-VII.0 for Fast Reactors,“ Nuclear
Data Evaluation Lab., Korea Atomic Energy Research Institute, Daejeon, 2008.

[14] M. B. Chadwick, P. Obložinský, M. Herman et al, „ENDF/B-VII.0: Next Generation

Evaluated Nuclear Data Library for Nuclear Science and Technology,“ Nuclear Data
Sheets, vol. 107, 1. issue.12, pp. 2931-3060, 2006.

[15] I. I. Bondarenko, Group Constants for Nuclear Reactor Calculations, Consultants

Bureau, 1964.

[16] G. I. Bell a S. Glasstone, Nuclear Reactor Theory, Litton Educational Publishing, 1970.

[17] R. S. Baker, „XREP Module,“ Los Alamos National Laboratory, 1993.

Proceedings of the International Conference Nuclear Energy for New Europe, Portorož, Slovenia, September 10 – 13, 2018