You are on page 1of 12

Annals of Nuclear Energy 96 (2016) 377–388

Contents lists available at ScienceDirect

Annals of Nuclear Energy


journal homepage: www.elsevier.com/locate/anucene

Monte Carlo analysis of the CROCUS benchmark on kinetics parameters


calculation
Andrea Zoia a,⇑, Yasushi Nauchi b, Emeric Brun a, Cédric Jouanne a
a
Den-Service d’Etudes des Réacteurs et de Mathématiques Appliquées (SERMA), CEA, Université Paris-Saclay, 91191 Gif-sur-Yvette, France
b
Central Research Institute of Electric Power Industry (CRIEPI), 2-11-1, Iwado Kita, Komae-shi, Tokyo 201-8511, Japan

a r t i c l e i n f o a b s t r a c t

Article history: The experimental configurations, the construction data and the material compositions of the CROCUS
Received 15 April 2016 reactor, a water-moderated zero-power core operated at the Ecole Polytechnique Fédérale de Lausanne
Received in revised form 17 June 2016 (EPFL), Switzerland, have been accurately determined and compiled into an OECD/NEA international
Accepted 19 June 2016
benchmark for reactivities and kinetics parameters calculations. As such, the CROCUS reactor configura-
Available online 7 July 2016
tions provide an outstanding testbed for validating Monte Carlo codes in the context of reactor physics. In
this work, we have performed extensive code–code comparisons between Monte carlo code TRIPOLI-4Ò
Keywords:
developed at CEA and the enhanced version of MCNP5 developed at CRIEPI based on MCNP5.1.30 on
CROCUS
Kinetics parameters
the whole set of CROCUS configurations available from OECD/NEA. First, we have verified the coherence
Monte Carlo of the respective independent Monte Carlo models and of the simulation options by computing reactiv-
TRIPOLI-4Ò ities from multiplication factors (the so-called direct reactivities). Then, we have calculated the kinetics
MCNP parameters by resorting to the Iterated Fission Probability (IFP) method and deduced the reactivities from
Iterated Fission Probability (IFP) the inhour equation (the so-called dynamic reactivities). An excellent agreement has been found between
the codes. Finally, we have carried out a detailed analysis of the simulation results and we have carefully
assessed the impact of nuclear data libraries on the computed quantities, with special emphasis on pos-
sible discrepancies between direct and dynamic reactivities.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction (Hoogenboom, 2003). To overcome this obstacle, two strategies


have been proposed so far: either resorting to perturbation tech-
In the context of reactor physics and criticality safety, niques (Verboomen et al., 2006; Nagaya and Mori, 2011), or finding
continuous-energy Monte Carlo simulation is considered as the an estimate of the adjoint neutron flux that can be computed dur-
gold standard to establish reference values for faster, but approxi- ing forward simulations. For this latter, the early attempts such as
mated, deterministic calculations (Spanier and Gelbard, 1969; Bell the method based on the expected number of fission neutrons in
and Glasstone, 1970). Monte Carlo codes are intrinsically based on the next generation (Nauchi and Kameyama, 2005) or the so-
the simulation of forward random walks: as such, they are ideally called Next Fission Probability approximation (Meulekamp and
suited to provide accurate estimates of physical observables van der Marck, 2006) have been shown to be possibly flawed for
weighted by the direct neutron flux. Un-weighted kinetics param- large and/or strongly heterogeneous reactors (Nagaya et al.,
eters can be straightforwardly obtained by running standard (for- 2010; Nauchi and Kameyama, 2010). A major breakthrough has
ward) criticality simulations: for instance, the mean generation been nonetheless made possible by the rediscovery (Feghhi et al.,
time K0 can be estimated by recording the neutron flight time from 2007, 2008; Nauchi and Kameyama, 2010; Kiedrowski et al.,
birth to fission, and the mean delayed neutron fraction b0 can be 2011) of the so-called Iterated Fission Probability (IFP) interpreta-
estimated by recording the fraction of delayed neutrons emitted tion of the adjoint flux (Soodak, 1949; Weinberg, 1952; Ussachoff,
at fission (Spriggs et al., 1997). 1955; Hurwitz, 1964).
Extending the capabilities of continuous-energy Monte Carlo Exact calculation of adjoint-weighted quantities by the IFP
codes to adjoint-weighted kinetics parameters demands the method thus establishes Monte Carlo simulation as a reference tool
adjoint neutron flux, which turns out to be a daunting task for the analysis of kinetics parameters (Nauchi and Kameyama,
2010; Kiedrowski et al., 2011; Shim et al., 2011). A number of pro-
⇑ Corresponding author. duction codes have integrated or are planning to integrate IFP
E-mail address: andrea.zoia@cea.fr (A. Zoia). capabilities: a non-exhaustive list includes MCNP (Kiedrowski,

http://dx.doi.org/10.1016/j.anucene.2016.06.024
0306-4549/Ó 2016 Elsevier Ltd. All rights reserved.
378 A. Zoia et al. / Annals of Nuclear Energy 96 (2016) 377–388

2011), SCALE (Perfetti, 2012), and SERPENT (Leppänen et al., of 100 W. The core is approximately cylindrical, with an outer
2014). Both CEA and CRIEPI have recently implemented methods diameter of about 60 cm and a height of 100 cm. The experimental
for computing adjoint-weighted and un-weighted kinetics param- configurations that have been compiled into the international
eters in their respective Monte Carlo codes, as detailed in Truchet OECD/NEA benchmark and that we have used for our Monte Carlo
et al. (2015) and Nauchi and Kameyama (2009). In particular, calculations are described in detail in OECD/NEA (2007) and
CEA develops the code TRIPOLI-4Ò (Brun et al., 2015), whereas CRIEPI Paratte et al. (2006). The fuel rods are of two kinds: the central
develops an enhanced version of MCNP5 based on MCNP5.1.30.1 zone is loaded with 336 uranium dioxide fuel rods at 1:806 wt%
The validation of such methods represents a major requirement enrichment in 235U, and the peripheral zone is loaded with 172
for Monte Carlo codes. The CROCUS reactor is a water-moderated or 176 metallic uranium rods at 0:947 wt% enrichment in 235U.
zero-power core operated at the Ecole Polytechnique Fédérale de The number and positions of the fuel rods is given in OECD/NEA
Lausanne (EPFL), Switzerland, for which detailed measurements (2007) and Paratte et al. (2006) and depends on the critical exper-
of the inverse reactor period have been reported for several iments (a radial view of the corresponding Monte Carlo model is
delayed-supercritical configurations (Früh, 1993). Two kinds of displayed in Fig. 1).
experiments were carried out, where reactivity was inserted by In the first set of experiments (core type A), the water moder-
either varying the water level within the core, or by extracting ator level was raised from the critical level H1 , making the reactor
an absorber rod (Boron or Erbium) initially located in the center delayed supercritical. The corresponding reactor period was deter-
of the core. The inserted reactivity can be estimated by directly mined for three different water levels H2 ; H3 and H4 . In the second
computing the effective multiplication factor, or by applying the set of experiments (core type B), the reactor was made critical by
inhour equation using the computed kinetics parameters and the inserting an absorber rod in the center of the core and then adjust-
experimental value of the inverse reactor period. The experimental ing the water level. Then, the reactor was made delayed supercrit-
configurations, the construction data and the material composi- ical by withdrawing the absorber rod and keeping the water level
tions have been accurately determined (including uncertainties) constant. The reactor period was measured for an absorber rod
and compiled into an OECD/NEA international benchmark (OECD/ containing water and natural Boron and for an absorber rod con-
NEA, 2007; Paratte et al., 2006). As such, the CROCUS reactor con- taining Erbium oxide pellets. The benchmark configurations are
figurations provide an outstanding testbed for validating reactivi- resumed in Table 1.
ties and kinetics parameters computed by Monte Carlo codes.2 The results of the first four participants to the CROCUS bench-
In this work, we will perform a joint code–code comparison of mark were published in 2006: the simulation were carried out
the reactivities and the kinetics parameters for the whole set of by using the 3d deterministic code HEXNOD with group constants
CROCUS reactor configurations available from the OECD/NEA from CASMO4 and libraries mainly taken from ENDF/B-IV and
benchmark. To this goal, CEA will run the development version ENDF/B-V, the 2d deterministic code HELIOS with the standard
of TRIPOLI-4Ò, based on release 4.10, and CRIEPI will run its enhanced 45-groups HELIOS library, the 2d deterministic code BOXER with
version of MCNP5 based on MCNP5.1.30. Our aim is threefold: first, a library based on JEF-1, and the Monte Carlo code MCU with a
we will assess the coherence of the respective independent Monte library based on ENDF/B-VI, JENDL-3.2, and BROND (OECD/NEA,
Carlo models and of the simulation options by computing and 2007; Paratte et al., 2006). These simulation results are reported
cross-checking direct reactivities from multiplication factors. Then, in OECD/NEA (2007) and Paratte et al. (2006): while direct reactiv-
we will calculate the kinetics parameters and deduce the dynamic ities from multiplication factors were found to be generally quite
reactivities from the inhour equation, which will allow testing the satisfactory, the agreement between the direct reactivities and
new methods for computing the kinetics parameters that have the dynamic reactivities inferred from the inhour equation on the
been independently implemented in our codes. Finally, we will basis of the computed kinetics parameters showed relatively large
carefully assess the impact of nuclear data on our simulation discrepancies for two of the participants. Possible lack of accuracy
results, by resorting to three different libraries, namely, JEFF- in the simulation methods and/or in the nuclear data chosen by
3.1.1 (Santamarina et al., 2009), ENDF/B-VII.0 (Chadwick, some of the participants were pointed out as a possible reason
2006) and JENDL-4.0 (Shibata et al., 2011). for the observed discrepancies in the calculated kinetics parame-
This paper is structured as follows: in Section 2, we briefly recall ters (OECD/NEA, 2007; Paratte et al., 2006).
the key features of the CROCUS reactor benchmark and summarize Concerning in particular Monte Carlo methods, the effective
the two sets of critical experiments that will be used for our vali- delayed neutron fraction was estimated in MCU by resorting to
dation purposes. In Section 3 we will provide an overview of the the so-called (approximate) prompt method (OECD/NEA, 2007;
Monte Carlo codes used for the simulation. Then, in Section 4 we Paratte et al., 2006), whereas no information is available on the
will present the numerical findings stemming from the code–code method used to estimate the effective mean generation time (based
comparison, and a discussion will be provided in Section 5. Conclu- on numerical values, we might conjecture that the computed quan-
sions will be finally drawn in Section 6. tity was actually the un-weighted mean generation time K0 ).
The CROCUS benchmark configurations have been later revis-
ited by several authors with different Monte Carlo codes, methods
2. The CROCUS reactor benchmark
and libraries. For instance, the effective delayed neutron fraction
was computed in Vollaire et al. (2006) and in Leppänen (2008)
The CROCUS reactor is a two-region uranium-fueled and light
by resorting to the Next Fission Probability approximation
water-moderated critical facility, with a maximum allowed power
(Meulekamp and van der Marck, 2006). In Nauchi and
1
Kameyama (2008), the effective delayed neutron fraction was
The IFP method has been implemented in the official release of MCNP in 2010.
determined by the method based on the expected number of fis-
However, on August 20th, 2014, Dr. Kiedrowski of LANL has reported a bug
concerning the IFP algorithm in MCNP5-1.60, MCNP6.1, and MCNP6.1.1 (see sion neutrons in the next generation (Nauchi and Kameyama,
https://mcnp.lanl.gov/BUGS/BUGS.shtml). Prior to them, Dr. Nauchi has inde- 2010). The impact of nuclear data libraries on the discrepancies
pendently implemented his own version of the IFP method for the enhanced MCNP5 between direct and dynamic reactivities was carefully discussed
version developed at CRIEPI in 2009 (Nauchi and Kameyama, 2009).
2
in Nauchi and Kameyama (2008). Un-weighted mean generation
We should mention here that intensive efforts are being made so as to produce
other benchmark-quality experimental results for the kinetics parameters of light
times have been computed in Leppänen (2008).
water reactors, such as in the case of the IPEN/MB-01 reactor (Santos and Diniz, To the best of our knowledge, no attempt has been made so far
2014). to perform a systematic Monte Carlo analysis of the reactivities and
A. Zoia et al. / Annals of Nuclear Energy 96 (2016) 377–388 379

fraction and mean generation time are now available (Truchet


et al., 2015). For IFP, a superposed-cycles implementation has been
chosen, with an arbitrary number of latent generations M. As a par-
ticular case, choosing M ¼ 1 leads to weighting by the expected
number of fission neutrons in the next generation, which for the
sake of conciseness we will then denote by IFP1 in the following.
In the development version of the code that was used for the sim-
ulations reported in the following section, further capabilities
allowing for partial delayed neutron fractions and precursor decay
constants (per family) have been added.

3.2. The enhanced MCNP5 code version at CRIEPI


Fig. 1. Radial cut of the TRIPOLI-4Ò model of the CROCUS core. Left. Type A core, used
for water level experiments (configurations H1 ; H2 ; H3 and H4 ). Right. Type B core, MCNP is one of the most representative Monte Carlo simulation
used for absorber rod experiments (configurations B0 ; B ; Er 0 and Er  ). codes for neutron transport, and is developed by Los Alamos
National Laboratory. The code is used as the reference for criticality
calculations with continuous-energy cross section data libraries in
Table 1
ACE format. At CRIEPI, calculation techniques for the reactor kinet-
The CROCUS benchmark configurations. Configurations H1 ; H2 ; H3 and H4 correspond
to the water level experiments performed on the type A core, with H1 at the critical ics parameters have been conceived by utilizing the eigenvalue cal-
water level. Configurations Bc and B correspond to the Boron absorber rod culation scheme of MCNP. The first attempt in 2003 was to
experiments, and configurations Er c and Er  correspond to the Erbium absorber rod implement a routine into MCNP-4C (Briesmeister, 2000) so as to
experiments, both performed on the type B core. Subscript c refers to a critical calculate a delayed neutron fraction weighted by the expected
configuration with the absorber rod inserted, whereas subscript  denotes a delayed
number of fission neutrons in next generation, i.e., IFP1 (Nauchi
supercritical configuration with the absorber rod extracted. The measured inverse
reactor periods x are also reported. and Kameyama, 2003). Another routine for IFP1-weighted neutron
generation time had been also developed (Nauchi and Kameyama,
Name Core type x [s1]
2005). Then, the delayed neutron fraction repartition per precursor
H1 A 0 family and fission nuclide were calculated with IFP1 (Nauchi and
H2 A 0.013422
Kameyama, 2006, 2008). The IFP algorithm was finally imple-
H3 A 0.018180
H4 A 0.023392 mented into MCNP-5.1.30 so as to calculate the effective delayed
neutron fraction and the neutron generation time in 2009
Bc B 0
B B 0.012856 (Nauchi and Kameyama, 2009, 2010). The IFP routines were further
extended to the delayed neutron fraction repartition per precursor
Er c B 0
Er  B 0.032976 family and fission nuclide for the purposes of this work.
The effective delayed neutron fraction is estimated by tagging
each delayed neutron per nuclide and per precursor family at a
the kinetics parameters on the CROCUS benchmark configurations given generation g after convergence of the power iteration. The
based on the adjoint-weighted kinetic parameters. tag is then transmitted to all the descendant neutrons. In a later
generation M þ g þ 1, the descendant neutrons of the tagged
delayed are recorded and provide an estimate of the asymptotic
3. Overview of the Monte Carlo codes importance times the weight of the delayed neutrons. Based on
this score, the effective delayed neutron fraction is easily deduced.
3.1. The TRIPOLI-4 code at CEA For the calculation of the neutron generation time, the lifetime of
the neutron inducing fission is recorded instead.
TRIPOLI-4Ò is a general-purpose Monte Carlo radiation transport In the implementation of the IFP kinetics parameters in the
code developed at CEA and devoted to shielding, reactor physics MCNP5 version developed at CRIEPI, care was taken not to alter
with depletion, criticality safety and nuclear instrumentation the sampling sequence of nuclear reactions in the eigenvalue cal-
(Brun et al., 2015). TRIPOLI-4Ò is the reference Monte Carlo code for culations of the original code. That can be easily verified by check-
CEA (including laboratories and reactor facilities) and the French ing the number of sources and collisions in each generation as well
utility company EDF. It is also the reference code of the CRISTAL as the neutron multiplication factor. For this reason, neither biased
Criticality Safety package (Gomit et al., 2011) developed with IRSN source sampling nor non-analog treatment of the neutron trans-
and AREVA. The code offers both fixed-source and criticality simu- port have been adopted in the CRIEPI version of the IFP algorithm.
lation modes. Neutrons are simulated in the energy range from The implementation of IFP for parallel simulations has not been
20 MeV to 105 eV. Particle transport is performed in continuous- attempted, yet. Thus, the calculation burden (including computer
energy, and the necessary nuclear data (i.e., point-wise cross- time and size of files for fission sources and the recorded data for
sections, scattering kernels, secondary energy-angle distributions, IFP) are somewhat non-negligible. Research is ongoing on how to
secondary particle yields, fission spectra, and so on) are read by improve these points.
the code from any evaluation written in ENDF-6 format (McLane,
2004). The code can directly access files in ENDF and PENDF for- 3.3. Nuclear data libraries
mat. For the unresolved resonance range, probability tables (when
present and requested) are generated for TRIPOLI-4Ò by using the For our code–code comparisons, three different libraries have
CALENDF code (Sublet et al., 2011). been selected, namely, JEFF-3.1.1 (Santamarina et al., 2009),
Concerning kinetics parameters calculations, the method based ENDF/B-VII.0 (Chadwick, 2006) and JENDL-4.0 (Shibata et al.,
on the expected number of fission neutrons in the next generation 2011). To preserve coherence, all libraries have been processed at
was already available in TRIPOLI-4Ò, as discussed, for instance, in Lee CEA at 294 K, which is the closest temperature compatible with
and Hugot (2011). Starting from version 4.10, un-weighted kinet- the benchmark specifications. Concerning probability tables,
ics parameters and IFP-based adjoint-weighted delayed neutron TRIPOLI-4Ò resorts to the CALENDF code (Sublet et al., 2011), whereas
380 A. Zoia et al. / Annals of Nuclear Energy 96 (2016) 377–388

the simulations run at CRIEPI relied upon the standard treatment


available in MCNP5 (X-5 Monte Carlo Team, 2003). Concerning
thermal data for bound hydrogen in water, JEFF-3.1.1 and
ENDF/B-VII.0 libraries have Sða; bÞ at 294 K, while JENDL-4.0
has Sða; bÞ at 296 K. The impact of this difference in temperature
will be discussed in the following. Doppler broadening of elastic
scattering differential cross sections has been enforced by using
the standard SVT model. The DBRC model for resonant nuclides,
although available (Zoia et al., 2012), has not been used, since it
is not expected to have a major impact on reactivities and kinetics
parameters at low temperature. It is important to remark that the
6-group delayed neutron data parameters for the actinides in the
ENDF/B-VII.1 library have been restored back to those given in
the ENDF/B-VI.8 library (Chadwick, 2011). In particular, it has
been shown that the ENDF/B-VII.0 library leads to a 15% differ-
ence from rod-drop experimental results, and that the delayed
neutron data in ENDF/B-VI.8 library display a better agreement
with experimental measurements (Chadwick, 2011). The impact
Fig. 2. Effective multiplication factors keff for the CROCUS configurations as
of nuclear data libraries on the simulation results will be thor-
computed by MCNP5 (squares) and TRIPOLI-4Ò (stars) and different nuclear data
oughly discussed in the following. libraries. Blue squares and red stars: JEFF-3.1.1. Magenta squares and green stars:
ENDF/B-VII.0. Black squares and cyan stars: JENDL-4.0. (For interpretation of the
3.4. Simulation parameters references to color in this figure legend, the reader is referred to the web version of
this article.)

For the whole set of the CROCUS benchmark configurations, we


have run criticality simulations with 5  102 inactive cycles and Table 2
2  104 active cycles. The number of simulated neutrons per cycle Effective multiplication factors keff for the CROCUS configurations as obtained with
JEFF-3.1.1, together with the associated statistical uncertainties r. We have
was 105 . The attained convergence on keff is less than 2:4 pcm for denoted M5 as a shorthand for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò.
all calculations. Concerning IFP, we have taken M ¼ 20 latent
Case M5 keff r [105] T4 keff r [105]
generations.
H1 1.00251 1.4 1.00246 2.4
H2 1.00338 2.3 1.00336 2.4
4. Simulation results H3 1.00358 1.5 1.00356 2.3
H4 1.00379 1.1 1.00378 2.4
In this Section we will report the simulation results for the Bc 1.00244 2.0 1.00241 2.3
code–code comparison on the whole set of the CROCUS benchmark B 1.00327 1.6 1.00326 2.3
configurations. Erc 1.00239 1.3 1.00232 2.3
Er 1.00402 1.0 1.00396 2.3

4.1. The direct reactivities

As a preliminary and necessary step towards the code–code Table 3


Effective multiplication factors keff for the CROCUS configurations as obtained with
comparison on the kinetics parameters, we have initially verified ENDF/B-VII.0, together with the associated statistical uncertainties r. We have
that TRIPOLI-4Ò and the enhanced version of MCNP5 obtained consis- denoted M5 as a shorthand for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò.
tent results on the direct reactivities qk coming from effective mul-
Case M5 keff r [105] T4 keff r [105]
tiplication factor calculations. According to the benchmark
specifications (OECD/NEA, 2007; Paratte et al., 2006), we define H1 1.00253 1.2 1.00249 2.4
H2 1.00341 0.9 1.00335 2.4
the direct reactivities as
H3 1.00363 1.4 1.00360 2.4
H4 1.00382 2.0 1.00383 2.4
1 1
qk ¼ c   ; ð1Þ Bc 1.00247 1.9 1.00248 2.3
keff keff
B 1.00330 1.6 1.00329 2.3
c
where keff is the computed effective multiplication factor for the Erc 1.00244 0.6 1.00235 2.4
 Er 1.00403 2.1 1.00400 2.3
critical configuration, and keff is the computed effective multiplica-
tion factor for each delayed supercritical configuration obtained as a
perturbation of the corresponding critical configuration.
The aim of this analysis is to cross-check the code options and probably due to the intrinsic differences between the two codes
the consistency of the Monte Carlo models of the CROCUS reactor (for instance, different models for probability tables). Results for
independently developed at CEA and at CRIEPI. JEFF-3.1.1 and ENDF/B-VII.0 are quite close to each other,
Simulation results for the multiplication factors keff are visually whereas results for JENDL-4.0 lie about 40–50 pcm below those
displayed in Fig. 2, and the numerical values are provided in Table 2 of the other libraries. This is most probably due to the temperature
for the JEFF-3.1.1 library, Table 3 for the ENDF/B-VII.0 library, for thermal data for bound hydrogen in JENDL-4.0 being available
and Table 4 for the JENDL-4.0 library, respectively. As a general at 296 K, whereas the other reactor components are at 294 K. This
comment, an excellent agreement is found between the keff values effect will be discussed later.
computed by MCNP5 and those computed by TRIPOLI-4Ò. For all con- The results for the direct reactivities qk (determined with
figurations and all tested libraries, discrepancies are below 10 pcm, respect to the corresponding critical configuration) are displayed
and in most cases these discrepancies lie well within 3r of com- in Fig. 3, and the numerical values are provided in Table 5 for the
bined statistical uncertainty. The observed discrepancies are most JEFF-3.1.1 library, Table 6 for the ENDF/B-VII.0 library, and
A. Zoia et al. / Annals of Nuclear Energy 96 (2016) 377–388 381

Table 4 Table 5
Effective multiplication factors for the CROCUS configurations as obtained with Direct reactivities qk [pcm] for the CROCUS configurations as obtained with JEFF-
JENDL-4.0, together with the associated statistical uncertainties r. In the column 3.1.1, together with the associated statistical uncertainties r. We have denoted M5 as
marked with (⁄) we report the effective multiplication factors corresponding to a a shorthand for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò.
modified JENDL-4.0 library, where hydrogen (including the associated Sða; bÞ) was
taken from ENDF/B-VII.0 at temperature 294 K. We have denoted M5 as a shorthand Case M5 qk r T4 qk r
for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò. H2 86.7 2.7 88.9 3.3
H3 106.9 2.1 109.3 3.3
Case M5 keff r [105] T4 keff r [105] T4 keff (⁄) r [105]
H4 127.6 1.8 130.9 3.3
H1 1.00211 1.6 1.00210 2.4 1.00256 2.3
B 81.9 2.6 84.4 3.3
H2 1.00302 1.6 1.00295 2.4 1.00345 2.4
H3 1.00321 1.2 1.00316 2.3 1.00367 2.4 Er 162.4 1.6 162.7 3.3
H4 1.00344 2.2 1.00344 2.3 1.00385 2.4
Bc 1.00206 1.4 1.00200 2.3 1.00251 2.4
B 1.00289 1.5 1.00283 2.3 1.00330 2.4
Er c 1.00201 2.4 1.00195 2.3 1.00247 2.3 Table 6
Er  1.00365 2.0 1.00362 2.3 1.00408 2.4 Direct reactivities qk [pcm] for the CROCUS configurations as obtained with ENDF/B-
VII.0, together with the associated statistical uncertainties r. We have denoted M5 as
a shorthand for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò.

Case M5 qk r T4 qk r
H2 87.5 1.5 85.6 3.3
H3 108.7 1.9 110.4 3.3
H4 128.2 2.3 133.0 3.3
B 82.3 2.5 80.8 3.3
Er 158.8 2.2 163.7 3.3

Table 7
Direct reactivities qk [pcm] for the CROCUS configurations as obtained with JENDL-
4.0, together with the associated statistical uncertainties r. In the column marked
with (⁄) we report the direct reactivities corresponding to a modified JENDL-4.0
library, where hydrogen (including the associated Sða; bÞ) was taken from ENDF/B-
VII.0 at temperature 294 K. We have denoted M5 as a shorthand for MCNP5 and T4 as
a shorthand for TRIPOLI-4Ò.

Case M5 qk r T4 qk r T4 qk (⁄) r
H2 90.1 2.2 84.8 3.3 88.6 3.3
H3 109.6 2.0 105.9 3.3 109.8 3.3
H4 131.9 2.7 133.4 3.3 128.3 3.3
Fig. 3. Direct reactivities qk [pcm] for the CROCUS configurations as computed by B 83.3 2.0 83.2 3.3 78.3 3.3
MCNP5 (squares) and TRIPOLI-4Ò (stars) and different nuclear data libraries. Blue
Er 162.7 3.0 165.7 3.3 160.4 3.3
squares and red stars: JEFF-3.1.1. Magenta squares and green stars: ENDF/B-
VII.0. Black squares and cyan stars: JENDL-4.0. (For interpretation of the
references to color in this figure legend, the reader is referred to the web version
of this article.)

Table 7 for the JENDL-4.0 library, respectively. Again, an excellent


agreement is found between MCNP5 and TRIPOLI-4Ò for the whole set
of configurations, and for all libraries. The temperature effect
occurring for JENDL-4.0 on the multiplication factors is absent
for the direct reactivities, which are obtained as differences of
1=keff .

4.2. The kinetics parameters

We turn now our attention to the analysis of kinetics parame-


ters. The simulation results for the effective delayed neutron frac-
tion beff obtained by resorting to IFP are displayed in Fig. 4 (top),
and the numerical values are provided in Table 8 for the JEFF-
3.1.1 library, Table 9 for the ENDF/B-VII.0 library, and Table 10
for the JENDL-4.0 library, respectively. The agreement between
the beff values computed by MCNP5 and those computed by
Fig. 4. Top. Effective delayed neutron fraction beff [pcm] for the CROCUS config-
TRIPOLI-4Ò is good for all configurations and all tested libraries. It
urations as computed by MCNP5 (squares) and TRIPOLI-4Ò (stars) with IFP and
is apparent from Fig. 4 that beff is almost constant for all configura- different nuclear data libraries. Blue squares and red stars: JEFF-3.1.1. Magenta
tions, for a given library, which is coherent with the reactivity squares and green stars: ENDF/B-VII.0. Black squares and cyan stars: JENDL-4.0.
insertion scheme in the CROCUS experiments only weakly affecting Bottom. Un-weighted delayed neutron fractions b0 [pcm] for the CROCUS config-
the fission rate ratio of 235U and 238U. There is however a system- urations as computed by MCNP5 (squares) and TRIPOLI-4Ò (stars) with IFP and
different nuclear data libraries. Blue squares and red stars: JEFF-3.1.1. Magenta
atic effect due to the nuclear data libraries: the average beff value squares and green stars: ENDF/B-VII.0. Black squares and cyan stars: JENDL-4.0.
for JEFF-3.1.1 is fairly larger than the average beff for JENDL- (For interpretation of the references to color in this figure legend, the reader is
4.0, which is in turn larger than the average beff for ENDF/B-VII.0. referred to the web version of this article.)
382 A. Zoia et al. / Annals of Nuclear Energy 96 (2016) 377–388

Table 8 ciable differences were found between the beff ;i values for the three
Effective delayed neutron fractions beff [pcm] for the CROCUS configurations as
critical configurations. The simulation results corresponding to
obtained with JEFF-3.1.1, together with the associated statistical uncertainties r.
We have denoted M5 as a shorthand for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò. ENDF/B-VII.0 and JENDL-4.0 are shown in Fig. 7 (bottom) for
the H1 configuration alone. Again, a very good agreement is found
Case M5 beff r T4 beff r M5 b0 r T4 b0 r between MCNP5 and TRIPOLI-4Ò for all families. For the first and sixth
H1 758.5 2.6 756.5 0.96 734.56 0.20 734.33 0.15 family, ENDF/B-VII.0 and JENDL-4.0 yield very similar values.
H2 759.6 3.1 758.7 0.98 734.39 0.20 734.47 0.15
For the second and third family ENDF/B-VII.0 yield a value smaller
H3 755.5 2.5 758.3 0.98 734.37 0.20 734.52 0.15
H4 761.9 2.2 758.5 0.97 734.30 0.20 734.40 0.15 than JENDL-4.0, and vice versa for the fourth and fifth family. The
numerical values are resumed in Table 14 for JEFF-3.1.1, Table 15
Bc 760.4 2.2 758.1 0.96 734.68 0.20 734.62 0.15
B 757.2 2.7 757.7 0.97 734.60 0.20 734.92 0.15 for ENDF/B-VII.0, and Table 16 for JENDL-4.0, respectively.
The un-weighted decay constants ki for JEFF-3.1.1 simply
Er c 752.6 2.3 759.0 0.97 734.60 0.20 734.70 0.15
Er  755.3 2.7 758.2 0.97 734.49 0.20 734.78 0.15 coincide with the data reported in the nuclear data library, since
all isotopes are associated to exactly the same precursor families.
For reference, the numerical values are resumed in Table 17. The
Table 9 computed ki for ENDF/B-VII.0 and JENDL-4.0 are displayed in
Effective delayed neutron fractions beff [pcm] for the CROCUS configurations as Fig. 7 for the H1 critical configuration. An excellent agreement is
obtained with ENDF/B-VII.0, together with the associated statistical uncertainties r. found between MCNP5 and TRIPOLI-4Ò for all families. The ki values
We have denoted M5 as a shorthand for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò. obtained with ENDF/B-VII.0 for the fifth and sixth family are sig-
Case M5 beff r T4 beff r M5 b0 r T4 b0 r nificantly larger than those obtained with JENDL-4.0. Numerical
H1 737.4 2.7 738.3 0.96 713.76 0.20 713.57 0.18
values are reported in Table 18 for ENDF/B-VII.0 and Table 19
H2 738.3 2.7 737.2 0.96 713.59 0.20 713.64 0.18 for JENDL-4.0, respectively.
H3 742.0 2.9 736.2 0.96 713.57 0.20 713.30 0.18
H4 741.4 2.9 737.2 0.96 713.55 0.20 714.00 0.18
4.4. The reactivities from the inhour equation
Bc 736.2 2.8 736.5 0.97 713.83 0.20 713.92 0.18
B 736.4 2.5 737.5 0.96 713.77 0.20 714.10 0.18
Based on the results discussed above, we have then computed
Er c 731.4 3.1 737.5 0.96 713.81 0.20 713.79 0.18
the dynamic reactivities qx stemming from the kinetics parame-
Er  739.4 3.0 734.9 0.95 713.64 0.20 713.69 0.18
ters, via the inhour equation
X x
qx ¼ xKeff þ beff;i ; ð2Þ
i
ki þ x
The simulation results for the effective mean generation time
Keff obtained by resorting to IFP are displayed in Fig. 5 (top), and where x is the inverse reactor period coming from experimental
the numerical values are provided in Table 11 for the JEFF- measurements (see Table 1), Keff is the computed effective mean
3.1.1 library, Table 12 for the ENDF/B-VII.0 library, and Table 13 generation time, ki are the computed decay constants per family i,
for the JENDL-4.0 library, respectively. The agreement between and beff;i are the computed effective delayed neutron fractions per
the Keff values computed by MCNP5 and those computed by TRI-
family i. According to the benchmark specifications (OECD/NEA,
Ò
POLI-4 is good for all configurations and all tested libraries. We 2007; Paratte et al., 2006), the kinetics parameters appearing in
can notice from Fig. 5 (top) that Keff is slightly but systematically Eq. (2) are those computed on the critical configurations.
higher for the Boron and Erbium absorber rod configurations (type Simulation results for qx are displayed in Fig. 8. It is apparent
B core), and decreases when increasing the water level (type A that for all configurations MCNP5 and TRIPOLI-4Ò are in very good
core) or when extracting the absorber rod (type B core). Similarly agreement. The impact of nuclear data libraries is quite prominent.
as in the case of beff , there is also a systematic effect due to the Indeed, JEFF-3.1.1 and JENDL-4.0 yield qx values that are con-
nuclear data libraries: the average Keff values for JEFF-3.1.1 sistent with each other and with the corresponding values of qk .
and JENDL-4.0 are close to each other, and lie above the average On the contrary, qx values for ENDF/B-VII.0 are systematically
Keff for ENDF/B-VII.0. Since Keff non-trivially depends on the and significantly lower than those obtained with the other two
interplay of several nuclear data components (including energy libraries, and are significantly lower than the corresponding values
spectra and angular distributions), a simple explanation of the of qk by about the same amount, namely,  15%. This issue might
observed trends could hardly be obtained. be related to the problems reported above concerning the applica-
bility of the ENDF/B-VII.0 library to the interpretation of reactor
4.3. Partial kinetics parameters per precursor family kinetics experiments. Numerical values are provided in Table 20
for JEFF-3.1.1, Table 21 for ENDF/B-VII.0, and Table 22 for
In view of the calculation of the dynamic reactivity qx , we have JENDL-4.0, respectively.
separately computed the IFP-weighted contributions beff ;i per Similar discrepancies between the computed direct and
precursor family and the un-weighted decay constants ki .3 The dynamic reactivities have been reported, e.g., in OECD/NEA
JEFF-3.1.1 has been recently updated with all fission products (2007), Paratte et al. (2006), Nauchi and Kameyama (2008) and
being associated to the same 8 precursor families. Since ENDF/B- Leppänen (2008). In these works, it was conjectured that the differ-
VII.0 and JENDL-4.0 are still using the 6 precursor families format ences between qk and qx were due to the calculation method,
(with possibly different decay constants), the analysis of JEFF-3.1.1 including the effects of the numerical methods chosen for the esti-
will be carried out separately. In Fig. 6 (top) we display the beff ;i val- mation of the kinetics parameters (for instance, the number of
ues for the three critical configurations H1 ; Bc and Erc as computed by energy groups for deterministic codes, or the use of approximated
resorting to JEFF-3.1.1. An excellent agreement is found between techniques such as the prompt method) and the effects of the
MCNP5 and TRIPOLI-4Ò for all families and all configurations. No appre- nuclear data libraries.
Based on our previous considerations, we can safely state that
3
The ki values are computed by weighting the decay constants retrieved in the
the CROCUS benchmark models used for our Monte Carlo simula-
nuclear data libraries by the (normalized) direct fission operator, as customary in the tions are consistent, and that the reference IFP methods used in
derivation of the point kinetics equations. our respective codes are correctly implemented.
A. Zoia et al. / Annals of Nuclear Energy 96 (2016) 377–388 383

Table 10
Effective delayed neutron fractions beff [pcm] for the CROCUS configurations as obtained with JENDL-4.0, together with the associated statistical uncertainties r. In the column
marked with (⁄) we report the effective delayed neutron fractions corresponding to a modified JENDL-4.0 library, where hydrogen (including the associated Sða; bÞ) was taken
from ENDF/B-VII.0 at temperature 294 K. We have denoted M5 as a shorthand for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò.

Case M5 beff r T4 beff r T4 b0 r T4 b0 r T4 beff (⁄) r


H1 743.1 2.8 742.6 0.97 720.3 0.20 720.1 0.18 744.1 0.96
H2 741.2 3.1 742.4 0.97 720.1 0.20 720.1 0.18 744.5 0.96
H3 743.5 2.4 742.2 0.95 720.1 0.20 719.9 0.18 743.4 0.96
H4 744.0 2.4 743.7 0.96 720.1 0.20 720.3 0.18 743.7 0.97
Bc 744.8 2.6 744.2 0.97 720.4 0.20 720.4 0.18 744.5 0.96
B 747.4 2.8 743.2 0.96 720.4 0.20 720.5 0.18 743.3 0.96
Er c 742.3 2.7 744.3 0.96 720.4 0.20 720.2 0.18 743.1 0.95
Er  745.7 3.0 742.4 0.96 720.2 0.20 720.2 0.18 742.0 0.96

Table 12
Effective mean generation time Keff [ls] for the CROCUS configurations as obtained
with ENDF/B-VII.0, together with the associated statistical uncertainties r. We have
denoted M5 as a shorthand for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò.

Case M5 Keff r T4 Keff r M5 K0 r T4 K0 r


H1 47.49 0.02 47.52 0.006 59.16 0.0057 59.18 0.0026
H2 47.43 0.02 47.50 0.006 59.17 0.0056 59.21 0.0026
H3 47.43 0.02 47.47 0.006 59.19 0.0057 59.22 0.0026
H4 47.43 0.02 47.47 0.006 59.20 0.0060 59.22 0.0026
Bc 47.61 0.03 47.61 0.006 59.11 0.0054 59.13 0.0025
B 47.52 0.02 47.62 0.006 59.05 0.0059 59.07 0.0025
Erc 47.55 0.02 47.62 0.006 59.19 0.0056 59.23 0.0026
Er 47.53 0.02 47.57 0.006 59.08 0.0055 59.10 0.0025

Paratte et al., 2006). Decay constants ki only differ in the fifth


and sixth family. Therefore, it is reasonable to conclude that the
key impact on qx is principally due to the discrepancies on the
Fig. 5. Top. Effective mean generation time Keff [ls] for the CROCUS configurations effective delayed neutron fraction beff and on the repartition beff;i .
as computed by MCNP5 (squares) and TRIPOLI-4Ò (stars) with IFP and different nuclear
data libraries. Blue squares and red stars: JEFF-3.1.1. Magenta squares and green
stars: ENDF/B-VII.0. Black squares and cyan stars: JENDL-4.0. Bottom. Un- 5. Discussion
weighted mean generation times K0 [ls] for the CROCUS configurations as
computed by MCNP5 (squares) and TRIPOLI-4Ò (stars) with IFP and different nuclear
data libraries. Blue squares and red stars: JEFF-3.1.1. Magenta squares and green
In this Section we provide some further considerations concern-
stars: ENDF/B-VII.0. Black squares and cyan stars: JENDL-4.0. (For interpretation ing the calculation of the kinetics parameters and their impact on
of the references to color in this figure legend, the reader is referred to the web the dynamic reactivities.
version of this article.)

5.1. Analysis of the delayed neutron fractions and reactivity


Table 11
components
Effective mean generation time Keff [ls] for the CROCUS configurations as obtained
with JEFF-3.1.1, together with the associated statistical uncertainties r. We have The enhanced MCNP5 code developed at CRIEPI has the capabil-
denoted M5 as a shorthand for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò. ity of separately determining the contribution of each nuclide to
Case M5 Keff r T4 Keff r M5 K0 r T4 K0 r the IFP-weighted kinetics parameters. The components of the nor-
malized delayed neutron fractions beff;i;j =beff per nuclide and per
H1 47.52 0.02 47.60 0.006 59.11 0.0054 59.14 0.0025
H2 47.54 0.03 47.65 0.006 59.14 0.0053 59.18 0.0025 family are displayed in Fig. 9 for the critical H1 configuration. The
H3 47.53 0.03 47.56 0.006 59.16 0.0054 59.19 0.0026 contributions of 235U are much stronger than those of 238U, as
H4 47.49 0.03 47.55 0.006 59.15 0.0053 59.19 0.0025 expected on physical grounds. The dominant family for both 235U
Bc 47.62 0.02 47.69 0.006 59.09 0.0059 59.11 0.0025 and 238U is the fifth when using JEFF-3.1.1 (Fig. 9, top), and the
B 47.66 0.02 47.70 0.006 59.03 0.0054 59.05 0.0025 fourth when using ENDF/B-VII.0 and JENDL-4.0 (Fig. 9, bottom).
Er c 47.65 0.03 47.68 0.006 59.17 0.0055 59.20 0.0026 Based on these results, it is also possible to compute the distinct
Er  47.59 0.03 47.66 0.006 59.04 0.0054 59.08 0.0025 contributions to the dynamic reactivities per nuclide and per fam-
ily, which are displayed in Fig. 10 for the H2 configuration. The con-
tributions of 235U to qx are again much stronger than those of 238U.
Thus, the observed discrepancies between qk and qx must pri- It is immediately apparent that the dominant contribution to qx is
marily stem from nuclear data libraries. Indeed, as shown in the given by the beff;i;j x=ðki þ xÞ term corresponding to the second
previous sections, beff and Keff are significantly different between family, for both JEFF-3.1.1 (Fig. 10, top), ENDF/B-VII.0 and
libraries, and the repartition of beff;i and ki also varies. The JEFF- JENDL-4.0 (Fig. 10, bottom). The shift between the dominant fam-
3.1.1 and JENDL-4.0 libraries provide consistent values with each ily for the delayed neutron fractions components and for the
other, so it is easier to compare JENDL-4.0 and ENDF/B-VII.0, dynamic reactivity must clearly be due to the weighting factor
which are based on the same number of families. Since the numer- f i ðxÞ ¼ x=ðki þ xÞ. In principle, f i ðxÞ depends on the inverse reac-
ical values of Keff are very small, discrepancies on the mean gener- tor period x, but it turns out that x is rather small for the range of
ation time are not expected to play a major role on the dynamic reactivity insertions explored in the CROCUS benchmark configura-
reactivity value (see, e.g., the discussion in OECD/NEA, 2007; tions, so that the contribution of the first few families are
384 A. Zoia et al. / Annals of Nuclear Energy 96 (2016) 377–388

Table 13
Effective mean generation time Keff [ls] for the CROCUS configurations as obtained with JENDL-4.0, together with the associated statistical uncertainties r. In the column
marked with (⁄) we report the effective delayed neutron fractions corresponding to a modified JENDL-4.0 library, where hydrogen (including the associated Sða; bÞ) was taken
from ENDF/B-VII.0 at temperature 294 K. We have denoted M5 as a shorthand for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò.

Case M5 Keff r T4 Keff r M5 K0 r T4 K0 r T4 Keff (⁄) r


H1 47.60 0.03 47.60 0.006 59.21 0.0053 59.24 0.0025 47.55 0.0060
H2 47.57 0.03 47.56 0.006 59.25 0.0049 59.27 0.0026 47.51 0.0059
H3 47.55 0.02 47.58 0.006 59.25 0.0055 59.28 0.0025 47.52 0.0059
H4 47.54 0.02 47.55 0.006 59.27 0.0053 59.28 0.0025 47.50 0.0060
Bc 47.68 0.03 47.69 0.006 59.18 0.0054 59.20 0.0025 47.65 0.0060
B 47.66 0.03 47.69 0.006 59.13 0.0052 59.14 0.0025 47.63 0.0060
Er c 47.67 0.03 47.70 0.006 59.26 0.0056 59.29 0.0026 47.65 0.0059
Er  47.66 0.03 47.66 0.006 59.14 0.0054 59.16 0.0025 47.60 0.0060

Table 14
Partial effective delayed neutron fractions beff;i [pcm] for the CROCUS configuration H1
as obtained with JEFF-3.1.1, together with the associated statistical uncertainties r.
We have denoted M5 as a shorthand for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò.

Family M5 beff;i r T4 beff;i r


1 22.56 0.16 22.75 0.17
2 110.25 0.38 109.42 0.37
3 64.60 0.27 64.55 0.28
4 143.80 0.42 143.05 0.42
5 246.37 0.55 244.93 0.54
6 80.05 0.31 80.21 0.31
7 66.65 0.27 66.56 0.28
8 24.73 0.17 25.00 0.18

Table 15
Partial effective delayed neutron fractions beff;i [pcm] for the CROCUS configuration H1
as obtained with ENDF/B-VII.0, together with the associated statistical uncertain-
ties r. We have denoted M5 as a shorthand for MCNP5 and T4 as a shorthand for TRIPOLI-
Ò
Fig. 6. Top. Effective delayed neutron fractions beff;i [pcm] per precursor family i for 4 .
the critical CROCUS configurations. IFP estimates are obtained by resorting to
MCNP5 (squares) and TRIPOLI-4Ò (stars) with JEFF-3.1.1, including 8 precursor Family M5 beff;i r T4 beff;i r
families. Blue squares and red stars: the H1 configuration. Magenta squares and 1 21.42 0.16 21.47 0.16
green stars: the Bc configuration. Black squares and cyan stars: the Erc configura- 2 118.06 0.37 117.86 0.38
tion. Bottom. Effective delayed neutron fractions beff;i [pcm] per precursor family i 3 115.93 0.38 116.40 0.38
for the H1 CROCUS configuration. IFP estimates are obtained by resorting to MCNP5 4 337.39 0.63 337.54 0.65
(squares) and TRIPOLI-4Ò (stars) with ENDF/B-VII.0 and JENDL-4.0, each including 5 108.39 0.38 108.68 0.37
6 precursor families. Blue squares and red stars: ENDF/B-VII.0. Magenta squares 6 36.35 0.22 36.38 0.21
and green stars: JENDL-4.0. (For interpretation of the references to color in this
figure legend, the reader is referred to the web version of this article.)

Table 16
Partial effective delayed neutron fractions beff;i [pcm] for the CROCUS configuration H1
as obtained with JENDL-4.0, together with the associated statistical uncertainties r.
In the column marked with (⁄) we report the effective delayed neutron fractions
corresponding to a modified JENDL-4.0 library, where hydrogen (including the
associated Sða; bÞ) was taken from ENDF/B-VII.0 at temperature 294 K. We have
denoted M5 as a shorthand for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò.

Family M5 beff;i r T4 beff;i r T4 beff;i (⁄) r


1 22.68 0.16 22.37 0.17 22.71 0.17
2 154.18 0.43 153.22 0.43 153.44 0.44
3 143.67 0.41 143.63 0.43 143.52 0.43
4 289.56 0.58 290.99 0.61 292.14 0.60
5 96.78 0.33 97.86 0.35 97.65 0.35
6 34.84 0.21 34.50 0.21 34.66 0.21

enhanced. In order to achieve a better agreement between direct


and dynamic reactivities, an accurate evaluation of the distinct
beff;i;j components (which depend on the accuracy of the available
nuclear data per precursor family) is thus indispensable, together
with that of the total beff value.
Fig. 7. Decay constants ki [s1] per precursor family i for the H1 CROCUS
configuration. Monte Carlo estimates are obtained by resorting to MCNP5 (squares) 5.2. Convergence of IFP kinetics parameters
and TRIPOLI-4Ò (stars) with ENDF/B-VII.0 and JENDL-4.0, each including 6
precursor families. Blue squares and red stars: ENDF/B-VII.0. Magenta squares
and green stars: JENDL-4.0. (For interpretation of the references to color in this In our previous analysis, we have exclusively focused on the
figure legend, the reader is referred to the web version of this article.) IFP-weighted kinetics parameters as obtained at convergence, i.e.,
A. Zoia et al. / Annals of Nuclear Energy 96 (2016) 377–388 385

Table 17
Decay constants ki [s1] from JEFF-3.1.1.

1 2 3 4 5 6 7 8
ki 0.012467 0.028292 0.042524 0.13304 0.292467 0.666488 1.63478 3.5546

Table 18 Table 20
Decay constants ki [s1] for the CROCUS configuration H1 as obtained with ENDF/B- Dynamic reactivities qx [pcm] for the CROCUS configurations as obtained with JEFF-
VII.0, together with the associated statistical uncertainties r. We have denoted M5 as 3.1.1, together with the associated statistical uncertainties r. We have denoted M5 as
a shorthand for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò. a shorthand for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò.

Family M5 ki r T4 ki r Case M5 qx r T4 qx r
1 0.01249 0.00029 0.01249 0.00003 H2 88.9 0.18 88.6 0.18
2 0.03172 0.00029 0.03158 0.00001 H3 110.6 0.22 110.2 0.22
3 0.10980 0.00029 0.11039 0.00004 H4 131.2 0.25 130.7 0.25
4 0.31855 0.00029 0.32078 0.00001
B 85.8 0.17 86.1 0.18
5 1.35174 0.00029 1.34851 0.00002
6 8.72204 0.00029 8.84426 0.00001 Er 162.3 0.29 162.6 0.30

Table 19 Table 21
Decay constants keff ;i [s1] for the CROCUS configuration H1 as obtained with JENDL- Dynamic reactivities qx [pcm] for the CROCUS configurations as obtained with ENDF/
4.0, together with the associated statistical uncertainties r. In the column marked B-VII.0, together with the associated statistical uncertainties r. We have denoted M5
with (⁄) we report the effective delayed neutron fractions corresponding to a as a shorthand for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò.
modified JENDL-4.0 library, where hydrogen (including the associated Sða; bÞ) was
taken from ENDF/B-VII.0 at temperature 294 K. We have denoted M5 as a shorthand Case M5 qx r T4 qx r
for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò. H2 73.7 0.16 73.6 0.16
H3 92.0 0.19 92.0 0.20
Family M5 ki r T4 ki r T4 ki (⁄) r
H4 109.6 0.22 109.5 0.23
1 0.01249 0.00029 0.01257 0.00004 0.01257 0.00004
B 71.2 0.16 71.1 0.16
2 0.03064 0.00029 0.03079 0.00003 0.03079 0.00003
3 0.11313 0.00029 0.11582 0.00002 0.11582 0.00002 Er 136.5 0.26 136.3 0.27
4 0.30507 0.00029 0.31077 0.00003 0.31077 0.00003
5 1.15374 0.00029 1.18130 0.00003 1.18129 0.00003
6 3.07859 0.00029 3.17896 0.00002 3.17892 0.00002
3.1.1 nuclear data library. For the sake of clarity, we have normal-
ized the values of beff and Keff to their asymptotic estimates
obtained at the maximum computed M. As a general remark, the
IFP-weighted delayed neutron fraction progressively increases as
a function of the latent generations, whereas the IFP-weighted
mean generation time decreases. The ratio between the asymptotic
IFP values and the IFP1 values at M ¼ 1 (the so-called generational
effect) is of the order of  4% for Keff and  2% for beff . For both
kinetics parameters, convergence to the asymptotic IFP-weighted
values is achieved within about 10 latent generations.

5.3. Convergence of the dynamic reactivity

Based on the convergence analysis of the IFP-weighted kinetics


parameters, it was also possible to assess the evolution of the
dynamic reactivity qx (which depends on the kinetics parameters
via the inhour equation) as a function of the latent generations M.
An example is shown in Fig. 12 for the critical H1 configuration
with the JEFF-3.1.1 nuclear data library. As expected, the conver-
gence is attained within the same number of generations as for beff
and Keff . Remark however that the divergence of the fluctuations
around the asymptotic value is stronger for qx : since the value
Fig. 8. Dynamic reactivities qx [pcm] for the CROCUS configurations as computed
by MCNP5 (squares) and TRIPOLI-4Ò (stars) and different nuclear data libraries. Blue of the dynamic reactivity is basically dominated by a small number
squares and red stars: JEFF-3.1.1. Magenta squares and green stars: ENDF/B- of precursors, statistical fluctuations appear significantly earlier
VII.0. Black squares and cyan stars: JENDL-4.0. (For interpretation of the than in the case of the kinetics parameters beff and Keff .
references to color in this figure legend, the reader is referred to the web version
of this article.)
5.4. Un-weighted kinetics parameters

after a large number of latent generations M. Both the enhanced As part of the code–code comparison, we have also computed
version of MCNP5 and TRIPOLI-4Ò can compute the evolution of beff the un-weighted kinetics parameters, i.e., the delayed neutron
and Keff as a function of the number of latent generations, due to fraction b0 and the mean neutron generation time K0 , for the whole
the convergence of the IFP estimates. This analysis has been carried set of CROCUS configurations, and for different libraries. The
out on the whole set of configurations: since convergence patterns simulation results are displayed in Fig. 4 (bottom) for b0 and
are pretty similar to each other, in Fig. 11 we only display the Fig. 5 (bottom) for K0 , respectively. For both parameters, an
results concerning the critical H1 configuration with the JEFF- excellent agreement is found between MCNP5 and TRIPOLI-4Ò. The
386 A. Zoia et al. / Annals of Nuclear Energy 96 (2016) 377–388

Table 22
Dynamic reactivities qx [pcm] for the CROCUS configurations as obtained with
JENDL-4.0, together with the associated statistical uncertainties r. In the column
marked with (⁄) we report the effective delayed neutron fractions corresponding to a
modified JENDL-4.0 library, where hydrogen (including the associated Sða; bÞ) was
taken from ENDF/B-VII.0 at temperature 294 K. We have denoted M5 as a shorthand
for MCNP5 and T4 as a shorthand for TRIPOLI-4Ò.

Case M5 qx r T4 qx r T4 qx (⁄) r
H2 87.5 0.18 86.3 0.18 86.6 0.18
H3 108.8 0.22 107.4 0.22 107.8 0.22
H4 129.1 0.25 127.5 0.25 127.8 0.24
B 84.7 0.18 84.0 0.18 83.9 0.19
Er  160.1 0.30 159.1 0.30 158.6 0.31

Fig. 11. Convergence of the effective delayed neutron fraction beff (blue) and of the
effective neutron generation time Keff (red) as a function of the number of latent
generations M. Both kinetics parameters have been normalized to their asymptotic
IFP-weighted values, evaluated at M ¼ 20. Simulation results are presented for the
critical H1 CROCUS configuration by resorting to the JEFF-3.1.1 nuclear data
library. Monte Carlo estimates are obtained by MCNP5 (squares) and TRIPOLI-4Ò
(stars). (For interpretation of the references to color in this figure legend, the reader
is referred to the web version of this article.)

un-weighted delayed neutron fractions are systematically smaller


(by about 3%) than the corresponding IFP-weighted beff . More-
over, similarly as in the case of beff ; b0 computed with JEFF-
3.1.1 is larger than those computed with JENDL-4.0, which in
turn are larger than those computed with ENDF/B-VII.0. The
un-weighted delayed neutron fraction is approximately constant
for all benchmark configurations. Observe that the difference
Fig. 9. Normalized components beff;i;j =beff of the effective delayed neutron fraction between un-weighted and IFP-weighted delayed neutron fractions
(per nuclide j and family i) for the critical H1 CROCUS configuration as computed by is not larger than the difference due to data libraries. Concerning
MCNP5. Top. Simulation results for the JEFF-3.1.1 library. Bottom. Simulation
K0 , un-weighted neutron generation times are systematically and
results for the ENDF/B-VII.0 library (blue squares) and for the JENDL-4.0 library
(red stars). (For interpretation of the references to color in this figure legend, the considerably larger than the corresponding IFP-weighted Keff .
reader is referred to the web version of this article.) The values of K0 estimated in type A cores increase with increas-
ing water level, which is expected on physical grounds. In type B
cores, K0 decreases when extracting the central absorber rod and
keeping the water level constant. When the absorber rod is
extracted, the peak of the thermal neutron flux lies at the center
of the core, due to spatial symmetries. When the absorber rod is
inserted, the thermal flux at the center of the core is depressed,
and the peak is shifted outward: the leakage of the thermal (i.e.,
longest-lived) neutrons is increased, which means that a higher
fraction of neutrons has spent time in the moderator, and the life-
time has also increased. Similarly as in the case of Keff , the K0 com-
puted with JENDL-4.0 are larger than those computed with ENDF/
B-VII.0, which in turn are larger than those computed with
JEFF-3.1.1.

5.5. The effects of temperature

As we mentioned in the previous sections, the Sða; bÞ thermal


data for bound hydrogen in water for the JENDL-4.0 library are
available at temperature 296 K, which is in contrast with the other
reactor components being at temperature 294 K. This leads to an
appreciable difference in the absolute keff values. In order to fully
Fig. 10. Normalized components qx;i;j =qx of the dynamic reactivity (per nuclide j assess the impact of the thermal data in the reactivity and kinetics
and family i) for the H2 CROCUS configuration as computed by MCNP5. Top. parameters calculations, we have performed additional simula-
Simulation results for the JEFF-3.1.1 library. Bottom. Simulation results for the tions using a modified JENDL-4.0 library where hydrogen (includ-
ENDF/B-VII.0 library (blue squares) and for the JENDL-4.0 library (red stars). The
ing the associated Sða; bÞ) was taken from ENDF/B-VII.0 at
component Lw corresponds to the Keff x term appearing in the inhour equation. (For
interpretation of the references to color in this figure legend, the reader is referred temperature 294 K. The corresponding computed keff are reported
to the web version of this article.) in Table 4: the values obtained by resorting to the modified
A. Zoia et al. / Annals of Nuclear Energy 96 (2016) 377–388 387

EPFL, whose experimental configurations, construction data and


material compositions are well documented and have been com-
piled into an international benchmark at OECD/NEA.
Preliminary work has consisted in computing effective multipli-
cation factors and direct reactivities for the whole set of CROCUS
configurations and for three different nuclear data libraries. An
excellent agreement was found between MCNP5 and TRIPOLI-4Ò,
which has allowed concluding that our respective Monte Carlo
models and simulation options were consistent with each other.
Then, we have focused on the IFP-weighted and un-weighted
kinetics parameters: simulation results as obtained with MCNP5
and TRIPOLI-4Ò were found to be in very good agreement, which
means that the Monte Carlo methods were correctly implemented
in our Monte Carlo codes.
Based on these findings, we could also analyse the impact of
nuclear data libraries on kinetics parameters. Actually, we have
been able to detect some systematic discrepancies between direct
reactivities, stemming from effective multiplication factors, and
Fig. 12. Convergence of the dynamic reactivity qx as a function of the number of
dynamic reactivities, obtained by injecting the computed kinetics
latent generations M. Simulation results are presented for the critical H1 CROCUS parameters into the inhour equation. In particular, the difference
configuration by resorting to the JEFF-3.1.1 nuclear data library. Monte Carlo between the direct and dynamic reactivity is up to 5 % for the cal-
estimates are obtained by MCNP5. culations using IFP-weighted delayed neutron fractions from
JEFF-3.1.1 and JENDL-4.0, and the ratio between the asymptotic
IFP values and the IFP1 values at M ¼ 1 is of the order of 2% for the
JENDL-4.0 library are larger by about  40 pcm than those com-
delayed neutron fraction and the dynamic reactivity. Future
puted with the standard JENDL-4.0 library and lie much closer
research work will be aimed at further investigating these results
to those obtained with JEFF-3.1.1 and ENDF/B-VII.0. The asso-
by addressing the perturbation of the kinetics parameters and
ciated direct reactivities are resumed in Table 7: since qk are
assessing the impact of the number of delayed neutrons as given
obtained from differences of (inverse) keff , no appreciable discrep-
in the nuclear data libraries.
ancies are found with respect to the case of the standard JENDL-
Finally, the present analysis of the CROCUS benchmark could be
4.0 library. Moreover, the impact of the modified Sða; bÞ on IFP-
extended to the direct estimation of the inverse reactor period by
weighted kinetics parameters (see Tables 10, 13 and 16) and on
resorting to the natural eigenvalues calculation methods. Such
decay constants (see Table 19) is almost negligible, due to the
techniques have been recently implemented in the enhanced ver-
small values of the discrepancies on keff . For the same reason, the sion of MCNP5 (Nauchi, 2014) and in TRIPOLI-4Ò (Zoia et al., 2014a,
impact on the dynamic reactivities (see Table 22) is also very b) and deserve to be fully verified and validated.
limited.

Acknowledgments
5.6. Comparison to the reactor noise experiments

TRIPOLIÒ and TRIPOLI-4Ò are registered trademarks of CEA. A.


To the best of our knowledge, direct experimental measure-
Zoia, E. Brun and C. Jouanne wish to thank AREVA and Électricité
ments concerning the kinetics parameters of type A and B cores
de France (EDF) for partial financial support.
are not available in literature. However, the decay constant at
delayed criticality aD ¼ beff =Keff has been recently obtained by
resorting to neutron noise analysis techniques for a CROCUS con- References
figuration very close to those reported in the benchmark specifica-
tions (Roland et al., 2013). Since the aD is rather insensitive to the
Bell, G.I., Glasstone, S., 1970. Nuclear Reactor Theory. Van Nostrand Reinhold
Company.
finer details of reactivity, it is possibly interesting to compare its Briesmeister, J.F. (Ed.), 2000. LA-13709-M.
experimental value to the Monte Carlo estimations obtained on Brun, E. et al., 2015. Ann. Nucl. Energy 82, 151–160.
Chadwick, M.B., 2006. Nucl. Data Sheets 107, 2931–3060.
the benchmark configurations. The measured value is reported to
Chadwick, M.B., 2011. Nucl. Data Sheets 112, 2887–2996.
be aD ’ 146:6  6:3 s1 (Roland et al., 2013), whereas our Monte dos Santos, A., Diniz, R., 2014. Nucl. Sci. Eng. 178, 459.
Carlo simulations yield aD ’ 158:9  0:2 s1 with JEFF-3.1.1, Feghhi, S.A.H., Shahriari, M., Afarideh, H., 2007. Ann. Nucl. Energy 34, 514.
aD ’ 155:4  0:2 s1 with ENDF/B-VII.0 and aD ’ 156:0 Feghhi, S.A.H., Shahriari, M., Afarideh, H., 2008. Ann. Nucl. Energy 35, 1397.
Früh, R., 1993. Réacteur CROCUS, complément au rapport de sécurité: réactivité et
0:2 s1 with JENDL-4.0, respectively, for the critical H1 configura- paramètres cinétiques. EPFL Lausanne, Switzerland, LPR 196.
tion 4. Gomit, J.M. et al., 2011. In: Proceedings of the ICNC 2011 conference, Edinburgh,
Scotland.
Hoogenboom, J.E., 2003. Nuc. Sci. Eng. 143, 99.
6. Conclusions Hurwitz Jr., H., 1964. In: Radkowsky, A. (Ed.), Naval Reactors Physics Handbook, vol.
I. Naval Reactors, Division of Reactor Development, United States Atomic
Energy Commission, p. 864.
The key goal of this work was to perform a code–code compar- Kiedrowski, B.C., 2011. Adjoint Weighting for Continuous-Energy Monte Carlo
ison between the enhanced version of MCNP5 developed at CRIEPI Radiation Transport (PhD thesis). Michigan University.
and TRIPOLI-4Ò developed at CEA, with special focus on the estima- Kiedrowski, B.C., Brown, F.B., Wilson, P.P.H., 2011. Nucl. Sci. Eng. 68, 226.
Lee, Y.K., Hugot, F.X., 2011. In: Proceedings of the M&C 2011 Conference, Rio de
tion of reactivities and kinetics parameters. To this aim, we have Janeiro, Brazil.
selected two series of cores of the CROCUS reactor operated at Leppänen, J., 2008. In: Proceedings of the IYNC2008 conference, Interlaken,
Switzerland.
Leppänen, J. et al., 2014. Ann. Nucl. Energy 65, 272.
4
It is worth noting that a MCNP simulation was also carried out in Roland et al. McLane, V., 2004. In: Brookhaven National Laboratory report BNL-NCS-44945-01/
(2013) on the exact Monte Carlo model of the reactor noise experiment, with the 04-REV KB0301042.
ENDF/B-VII.0 library: the computed aD reads 155:0  1:6 s1. Meulekamp, R.K., van der Marck, S.C., 2006. Nucl. Sci. Eng. 152, 142.
388 A. Zoia et al. / Annals of Nuclear Energy 96 (2016) 377–388

Nagaya, Y. et al., 2010. Ann. Nucl. Energy 37, 1308. Shibata, K., Iwamoto, O., Nakagawa, T., Iwamoto, N., Ichihara, A., Kunieda, S., Chiba,
Nagaya, Y., Mori, T., 2011. Ann. Nucl. Energy 38, 254. S., Furutaka, K., Otuka, N., Ohsawa, T., Murata, T., Matsunobu, H., Zukeran, A.,
Nauchi, Y., 2014. In: Proceedings of the SNA+MC 2013 conference, Paris, France. Kamada, S., Katakura, J., 2011. J. Nucl. Sci. Technol. 48, 1–30.
Nauchi, Y., Kameyama, T., 2003. In: Preprints of the 2003 annual meeting of Atomic Shim, H.J., Kim, C.H., Kim, Y., 2011. J. Nucl. Sci. Technol. 48, 1453.
Energy Society of Japan, Sasebo, Japan, March 27–29 H18 [in Japanese]. Soodak, H., 1949. Pile kinetics. In: Goodman, G. (Ed.), The Science and Engineering of
Nauchi, Y., Kameyama, T., 2005. J. Nucl. Sci. Technol. 42, 503. Nuclear Power, vol. II. Addison-Wesley Press Inc, Cambridge, pp. 89–102.
Nauchi, Y., Kameyama, T., 2006. In: Proceedings of the PHYSOR2006 conference, Spanier, J., Gelbard, E.M., 1969. Monte Carlo Principles and Neutron Transport
Vancouver, Canada. Problems. Addison-Wesley, Massachusetts, USA.
Nauchi, Y., Kameyama, T., 2008. In: Proceedings of the PHYSOR2008 conference, Spriggs, G.D. et al., 1997. LA-UR-97-1073. Los Alamos National Laboratory.
Interlaken, Switzerland. Sublet, J.C., et al., 2011. CALENDF-2010: User Manual, Rapport CEA-R-6277.
Nauchi, Y., Kameyama, T., 2009. In: Preprints of the 2009 annual meeting of Atomic Truchet, G., Leconte, P., Santamarina, A., Damian, F., Brun, E., Zoia, A., 2015. Ann.
Energy Society of Japan, Tokyo, Japan, March 23–25 K41 [in Japanese]. Nucl. Energy 85, 17.
Nauchi, Y., Kameyama, T., 2010. J. Nucl. Sci. Technol. 47, 977. Ussachoff, L.N., 1955. In: Proceedings of the 1st UN conference of the peaceful uses
OECD/NEA, 2007. Benchmark on the kinetics parameters of the CROCUS reactor, in of atomic energy, P/503, Geneva, Switzerland.
Plutonium recycling, vol. IX, NEA 4440. Verboomen, B., Haeck, W., Baeten, P., 2006. Ann. Nucl. Energy 33, 911.
Paratte, J.M., Früh, R., Kasemeyer, U., Kalugin, M.A., Timm, W., Chawla, R., 2006. Ann. Vollaire, J., Plaschy, M., Jatuff, F., Chawla, R., 2006. In: Proceedings of the
Nucl. Energy 33, 739. PHYSOR2006 conference, Vancouver, Canada.
Perfetti, C.M., 2012. Advanced Monte Carlo Methods for Eigenvalue Sensitivity Weinberg, A.M., 1952. Am. J. Phys. 20, 401.
Coefficient Calculations (PhD thesis). Michigan University. X-5 Monte Carlo Team, 2003. MCNP – A general Monte Carlo N-particle transport
Roland, V., Perret, G., Girardin, G., Frajtag, P., Pautz, A., 2013. In: Neutron Noise code, Version 5, Volume I: Overview and Theory. Los Alamos National
Measurement at CROCUS, from Joint IGORR 2013 and IAEA Technology Meeting, Laboratory, LA-UR-03-1987.
Daejeon, Korea. Zoia, A., Brun, E., Jouanne, C., Malvagi, F., 2012. Ann. Nucl. Energy 54, 218–226.
Santamarina, A. et al., 2009. The JEFF-3.1.1 Nuclear Data Library. OECD-NEA Data Zoia, A., Brun, E., Malvagi, F., 2014a. Ann. Nucl. Energy 63, 276.
Bank, JEFF Report 22. Zoia, A., Brun, E., Damian, F., Malvagi, F., 2014b. Ann. Nucl. Energy 75, 627.

You might also like