MCNPX Benchmark Tests of Neutron Production in Massive Lead Target

1
MCNPX Benchmark Tests of Neutron Production in
Massive Lead Target
Mitja Majerle1, J. Adam1, P. Caloun1, S. A.
Gustov2, V. Henzl1, D. Henzlová1, V. G.
Kalinnikov2, A. Krása1, M. I. Krivopustov2, F.
Kríek1, A. Kugler1, I. V. Mirokhin2, A. A.
Solnyshkin2, V. M. Tsoupko-Sitnikov2, V. Wagner1
) Electronic address majerle_at_ujf.cas.cz 1)
Nuclear Physics Institute of the Academy of
Science of the Czech Republic, Prague, The Czech
Republic 2) Joint Institute for Nuclear
Research, Dubna, Russia
The PHASOTRON experiment We used continuos,
intensive (1013 protons/s), stable beam of
protons with the energy 660 MeV from the Dubna
Phasotron. The protons were directed to a lead
target (cylinder with r4.8 cm and d4x12cm). The
target was placed in a long, narrow corridor,
bounded with concrete walls. A set of monitor
detectors (few mm thick Al and Cu foils) was
placed directly in the beam in front of the
target. Neutron detectors (Al, Au, and Bi foils)
were placed on top of the target, along its whole
length. After the irradiation, the activity of
the detectors was measured in HPGe detectors, and
the production rates of transmuted elements were
calculated.
Motivation ADS are future technology, the
combination of a classical reactor with an
accelerator. The basic principle is to produce a
large number of neutrons in the spallation
process (relativistic ions heavy metal target),
and to introduce them into a sub-critical reactor
assembly. Extra neutrons are used to produce fuel
from 232Th, and/or to transmute long-lived
nuclear waste to short-lived isotopes. At the
JINR, series of experiments with different
targets (lead, tungsten, bismuth cylinders,
surrounded with polyethylene or with natural
uranium and polyethylene) were performed. The
physical aim of the experiments was to study
nuclear processes that occur in the target, to
measure the transmutation rates for higher
actinides and fission products, to measure the
heat production, etc. The NAA (Neutron Activation
Analysis) was mostly used to measure the neutron
field and the transmutation rates. The
experimental results are used to test two
calculation codes DCM (Dubna Cascade Model) and
MCNPX (LAHET in combination with MCNP). This
paper is focused on the calculations with MCNPX
2.4.0 of the the Phasotron experiment.
Schematics of the setup
Experimental results (Al left, Au right)
Calculation principles Our calculations were
done with MCNPX 2.4.0. We described our setup as
a cylindrical lead target, to which a proton beam
is directed. To compare the calculations with the
experimental data, we had to calculate the
production rates. We tried two different methods.
Convolution of the neutron spectra with
cross-sections We record the neutrons that cross
a plane of interest - SSW (Surface Source Write)
card. We classify the neutrons at detector
positions in energy bins - HTAPE3X. The result is
the neutron spectra on the top of the target
along its lenght. The production rates in the
detectors are calculated from the neutron
spectrum as follows Fn (E) is the energy
dependent neutron flux, s(E) is the microscopic
cross-section at energy E for a specific
reaction, CNA/M is the normalization constant.
For neutrons in energy bins we use
Cross-sections were taken from ENDF library,
or were calculated from the experimental data
(EXFOR). Direct calculation of production
rates MCNPX can directly multiply each neutron
that crosses the detector with the microscopic
cross-section for a given reaction F4 card with
FM multiplier card. The direct method is faster,
the cross-sections are taken from la150n
library. The results have to be multiplied with
the normalization constant CNA/M, where NA is
Avogrados number and M is the molecular mass of
the detector material. Problem the results
from two methods are not always the same !
Calculated neutron spectra along the target
Calculated production rates
Preliminary results
Ratio between results from HTAPE3x and from F4
card methods.
Experimental and calculated (HTAPE3X) rates for Bi
The influence of the calculation parameters The
influence of the setup parts
The influence of the beam geometry The
accelerator beam had a Gaussian shape and was
displaced. Extensive simulations on this and
other setups taught us that the beam, displaced
for 3 mm results in a change of neutron field for
ca. 5. The influence of the different
intra-cascade models The simulations were done
using the intra-nuclear cascade model BERTINI
INC. The calculations with two other models - CEM
INC and ISABEL INC - showed that the choice of
the model does not influence the results. The
models describe the nuclear reactions the same in
the range of few hundreds MeV.
WITH WALLS
WITHOUT WALLS
Comparison with experimental data We found out
that we can simplify our simulations to a bare
lead target inside concrete corridor, a simple
simulation with BERTINI INC is sufficient.
Holders, beam tubes, etc. do not need to be taken
in account. Knowing the exact position of the
beam and its geometry is important.
Calculated neutron spectra for a bare target
(left) and for a target inside the corridor
(right)
We included concrete walls in our calculations.
They function as a neutron moderator and reflect
a significant part of neutrons back. The refleced
neutrons are homogeneously distributed along the
target lenght. These neutrons produce 198Au in a
non-threshold reaction 197Au(n,g)198Au. The iron
components do not influence the results
significantly. This is expected, iron causes only
scattering of neutrons, but our detectors lie on
place where these effects are minimal.
Calculated prduction rates for 198Au
(left) Comparison of calculations and
experimental data (right) the agreement is
somewhere inside 30
Conclusions We are testing the capabilities of
MCNPX, exploring which calculation methods are
the best for our setups and consequently ADS
system. A lot of experimental data from
experiments with similar setups waits to be
simulated. With MCNPX, we are able to describe
the systems similar to ADS with the accuracy of
ca. 30. Calculating in parallel with MCNPX
gives very good results for our setups the
speed almost linearly rises with the number of
used processors. Calculating in parallel with PVM
is now used at all our calculations. We want to
test the dependency of the speed of simulations
on the number of used processors and the number
of events to find an optimal number of computers
in case of building a computational farm for ADS
simulations. Acknowledgments The authors are
grateful to the staff of the Dubna Phasotron
accelerator for providing a good proton beam for
our experiment. These experiments were supported
by the Czech Committee for collaboration with
JINR Dubna. This work was carried out partly
under support of the Grant Agency of the Czech
Republic (grant No. 202/03/H043) and ASCR
K1048102 (the Czech Republic). References On
the field of ADS, the author most profited from
the articles of K.D. Tolstov, C.D. Bowman, C.
Rubbia. Data about the experiment comes from many
publications of co-authors and my work. A great
guide on how to build the cluster was found on
Echelon Beowulf Cluster homepage, and in Linux
how-tos. MCNPX manual and internet forums on
RSICC were the main resources when trying to get
MCNPX with PVM work and when discovering MCNPX
capabilities.
MCNPX in parallel Four computers from our office
with preinstalled systems can be temporarily
changed to computing workstations. We used
Slackware Linux 10.1 on the server, the Etherboot
method to boot hosts, NFS (Network File System)
for their file system, directories which need
writting permissions were mounted as ramdisks.
Hard-drives of hosts were not used. Using this
method, we can easily extend the cluster to a
random number of computers. The information on
how to build such a cluster was found on
internet. To calculate in parallel, a software
package that distributes the load on other
processors has to be used. MCNPX 2.4.0 has
built-in support for PVM (Parallel Virtual
Machine). PVM was installed on hosts, MCNPX was
compiled with the PVM support, and the efficiency
of the parallel computing was tested.
The dependancy of the speed of calcualtions on
the number of used processors for a heterogenous
cluster (up) and on 2 two-processor machines
(left)
We have two two-processor computers (processors
Xeon 2 GHz) with preinstalled Linux, intended for
calculations (MCNPX, ROOT,...). In setting up the
cluster, we gained experience that helped us to
set the machines intended for calculations to
work in parallel. They calculated faster than our
cluster. We tried MCNPX, compiled with
different compilers (PGF90, G95, GFORTRAN, Intel
Fortran Compiler). Intel Fortran Compiler did the
best job, MCNPX compiled with it calculates 40
faster than with PGF90.