You are on page 1of 16

Radiation Measurements 99 (2017) 25e40

Contents lists available at ScienceDirect

Radiation Measurements
journal homepage: www.elsevier.com/locate/radmeas

Secondary neutrons inside a proton therapy facility: MCNPX


simulations compared to measurements performed with a Bonner
Sphere Spectrometer and neutron H*(10) monitors
rie De Smet a, b, *, Marijke De Saint-Hubert c, Nesrine Dinar d, e,
Vale
Giacomo Paolo Manessi e, Eleni Aza e, f, Christopher Cassell g, Clarita Saldarriaga Vargas c,
Olivier Van Hoey c, Gilles Mathot h, Frede
ric Stichelbaut h, Gilles De Lentdecker b,
Isabelle Gerardy , Marco Silari , Filip Vanhavere c
a e

a
Institut de Recherche de l’Institut Sup
erieur Industriel de Bruxelles (IRISIB), D
epartement Nucl eaire, Rue Royale 150, 1000 Bruxelles, Belgium
b
Universite Libre de Bruxelles, Inter-University Institute for High Energies (IIHE), Boulevard du Triomphe CP230, 1050 Bruxelles, Belgium
c
SCKCEN, Belgian Nuclear Research Centre, Boeretang 200, 2400 Mol, Belgium
d
Universite Joseph Fourier, European Master of Radiation Protection, Grenoble, France
e
CERN, CH-1211 Geneva 23, Switzerland
f
AUTH, Department of Physics, 54124 Thessaloniki, Greece
g
Politecnico of Milan, Department of Energy, Via Ponzio 34/3, 20133 Milan, Italy
h
Ion Beam Applications S.A. (IBA), Chemin du Cyclotron 3, 1348 Louvain-la-Neuve, Belgium

h i g h l i g h t s

 WENDI-2 measures H*(10) with an accuracy within 10% inside and outside shielding.
 LB 6411 underestimates H*(10) by 25e40% at forward angle or outside the shielding.
 MCNPX simulations are conservative regarding wall-transmitted neutron fluxes.
 Bertini & Dresner models predict the global neutron production in water quite well.
 Bertini overestimates neutron production above 100 MeV in water at forward angle.

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

Article history: Neutron spectrometry measurements with an extended-range Bonner Sphere Spectrometer (BSS), as
Received 22 June 2015 well as neutron H*(10) measurements using an extended-range rem meter WENDI-2, a conventional rem
Received in revised form meter LB 6411 and a tissue-equivalent proportional counter, were performed inside and around the
28 November 2016
Fixed-Beam Treatment Room at the proton therapy facility of Essen, in Germany. The WENDI-2 stood out
Accepted 13 March 2017
Available online 16 March 2017
as the easiest detector for making accurate neutron H*(10) measurements, since its direct measurements
were equivalent to the H*(10) rates obtained with the BSS. The measurements were also compared to
simulation results obtained with MCNPX 2.7.0 using two different selections of physics models for the
Keywords:
Proton therapy
hadron interactions above 150 MeV: the Bertini & Dresner models and the CEM03 model. For neutron
Neutron H*(10) H*(10) rates outside the treatment room, factors of 1.6e1.8 were obtained between the results of the two
MCNPX simulations, the Bertini & Dresner models yielding the largest values in all positions. The comparison of
Rem counters the simulation results with the WENDI-2 and BSS measurements for positions inside the treatment room
Bonner Sphere Spectrometer showed that the Bertini & Dresner models reproduce the global neutron production in the water
TEPC phantom relatively well, whereas the CEM03 model underestimates it by a factor of ~1.3. At the most-
forward angle, however, the Bertini model (unlike the CEM03 model) seemed to overestimate the
production of neutrons with energies above 100 MeV. Outside the shielding, the simulated H*(10)
overestimated the WENDI-2 measurements by factors of 2e3 with the Bertini & Dresner models, and 1.1
e1.7 with the CEM03 model. Both simulations were thus conservative with respect to the neutron fluxes
transmitted through the concrete walls. This conservative behaviour is probably caused by a combination

 Catholique de Louvain (UCL), Institute of Information and Communication Technologies, Electronics and Applied
* Corresponding author. Present address:Universite
Mathematics (ICTEAM), ELEN - Place du Levant 2 bte L5.04.04, 1348 Louvain-la-Neuve, Belgium.
E-mail address: va.desmet@uclouvain.be (V. De Smet).

http://dx.doi.org/10.1016/j.radmeas.2017.03.005
1350-4487/© 2017 Elsevier Ltd. All rights reserved.
26 V. De Smet et al. / Radiation Measurements 99 (2017) 25e40

of several uncertainties, including for instance uncertainties on the proton and neutron interaction cross-
sections and uncertainties on the concrete composition and density.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction are generally also based on the highest deliverable proton energy,
because the average number of secondary neutrons produced per
During irradiations in proton therapy facilities, secondary neu- proton increases with the energy of the proton beam.
trons with energies up to typically ~230 MeV are produced and Despite the conservative choices, it remains important to assess
constitute the main radiation hazard for the professionally exposed the overall conservative nature of the original shielding design
workers and members of the public. Therefore, the accelerator simulations for existing proton therapy facilities. The simulation
room and the treatment rooms are surrounded by very thick con- method should indeed be validated experimentally by means of
crete walls (typically ~2 m thick), designed to attenuate the stray comparisons with on-site neutron measurements performed
radiation field. outside the shielding. The scientific literature however contains
The shielding design of proton therapy facilities is often based relatively few comparisons of this type. Two recent studies (De
on the results of simulations carried out with Monte Carlo (MC) Smet et al., 2014b; Satoh et al., 2012) compared WENDI-2 mea-
codes like MCNPX (Pelowitz, 2011), FLUKA (Bo € hlen et al., 2014; surements to MC simulations of the neutron H*(10) rate around
Ferrari et al., 2005) or PHITS (Sato et al., 2013). According to proton therapy rooms. They tended to confirm the conservative
benchmark experiments conducted under well-controlled condi- nature of these MC simulations, as the simulated H*(10) rates
tions, FLUKA and MCNPX allow reproducing with relatively good overestimated the measurements by at least a factor of 2 in all
accuracy (<15%e35%) the neutron ambient dose equivalent H*(10) positions located behind the shielding walls.
(ICRP, 2010) outside the shielding of experimental facilities such as In (De Smet et al., 2014b), the accuracy of the WENDI-2 mea-
Cave A at GSI (Rollet et al., 2009; Silari et al., 2009; Wiegel et al., surements in terms of H*(10) was however not explicitly assessed.
2009) and the CERN-EU Reference Field (CERF) (Caresana et al., The WENDI-2 has a good sensitivity over the entire width of the
2014; Mayer et al., 2007). neutron spectra (from thermal energies up to ~230 MeV), but the
In shielding design simulations of future proton therapy facil- level of accuracy achieved still depends on the local neutron
ities, however, several parameters tend to be difficult to control in a spectrum and the corresponding adequacy of the detector response
precise way. The exact average concrete composition of the future function and calibration factor. An ideal detector of which the dose
building is for instance rarely known in advance. Often, only a response function perfectly reproduces the fluence-to-H*(10) con-
minimum density is selected beforehand and imposed on the version coefficients (ICRP, 1996; Sannikov and Savitskaya, 1997)
concrete provider. Also, the geometry of the treatment rooms or the does indeed not exist. Since the neutron spectra inside proton
accelerator room is rather complex to model due to, for instance, therapy facilities differ quite a lot from the 252Cf spectrum with
the presence of equipment that may have a significant influence on which the WENDI-2 is usually calibrated, it is advisable to inves-
the distribution of the neutron fluxes (beam line components, tigate whether field-specific calibration correction factors should
treatment nozzles, gantries, etc.). Typically, only a limited amount be applied to WENDI-2 measurements performed inside proton
of detail is represented in the modelled geometry because of time €gerhofer et al. inferred, based on an experiment
therapy facilities. Ja
constraints. Informed physical insight is thus required to identify involving quasi-monoenergetic neutron beams of 246 MeV and
the most important elements to model. In some cases, it may also 389 MeV at the RCNP of Osaka, that it would be advisable to include
not be straightforward to correctly model the proton beam losses a measurement uncertainty of 20% for WENDI-2 measurements
(inside the accelerator, for example). Apart from these issues, MC (with the standard 252Cf calibration) performed behind the
simulations are also affected by uncertainties on the used cross- shielding of primary proton beams of ~250 MeV (J€ agerhofer et al.,
section data for the interactions of protons and neutrons with 2011). The WENDI-2 is thus expected to provide reasonably accu-
matter. These uncertainties tend to be larger in the case of pro- rate neutron H*(10) measurements inside a proton therapy facility,
jectile energies above ~20 MeV since experimental data are un- but the achievable accuracy should nevertheless be formally
fortunately more scarcely available at high energies. Where assessed.
possible, libraries of evaluated cross-section data (based on This paper presents an intercomparison carried out around the
experimental data) are preferably used in the simulations. Above Fixed-Beam Treatment Room (FBTR) of the Westdeutsches Proto-
the upper limit of the evaluated data tables, however, theoretical nentherapiezentrum Essen, in Germany. Neutron H*(10) measure-
physics models need to be invoked (e.g. above 150 MeV with ments were performed with a WENDI-2 in several positions inside
MCNPX). Different hadronic models have been developed over the and outside the treatment room. In four of these positions,
past decades but none of them truly stands out as the most accurate including one position outside the treatment room, the neutron
in general. It is thus up to the user to assess which physics models energy spectrum was measured using a Bonner Sphere Spectrom-
ought to be the most suitable to the considered case. eter (BSS). A few additional comparisons were made with mea-
As a consequence, a conservative approach is preferably adopted surements from a Berthold LB 6411 rem counter, a tissue-equivalent
in MC shielding design simulations for proton therapy facilities. In a proportional counter (TEPC) and a bare 3He proportional counter.
conservative simulation, the neutron doses calculated outside the All of the measurements are compared to simulations of the
shielding are in principle overestimated, so that a safety margin will FBTR performed with MCNPX 2.7.0 in a similar way as in a previous
be implicitly included in the shielding design. A typical conserva- study (De Smet et al., 2014b). This time, not only the H*(10) rates
tive choice consists for instance in defining a concrete density that were simulated but also the WENDI-2 responses (see section 2.6.2),
is slightly lower than the minimum density requested to the con- in order to investigate e in this case, through simulation e the
crete provider. Another one consists in selecting the physics models accuracy of the WENDI-2 responses and the potential need for
that lead to the highest calculated neutron doses. The simulations calibration correction factors. As the simulated neutron spectra
V. De Smet et al. / Radiation Measurements 99 (2017) 25e40 27

might however not be perfectly realistic, this aspect was also


assessed experimentally by comparing the WENDI-2 measure-
ments to the H*(10) rates derived from the neutron spectra
measured by the BSS. These H*(10) measurements were also
checked for consistency with the TEPC measurements, which rely
on microdosimetric principles (see section 2.4).
Another main goal of this work was to investigate the impact of
the selected physics models (available within MCNPX) on the
simulated neutron spectra and H*(10) rates. As will be explained in
section 2.6, two versions of the same simulation are compared in
this paper: one using, as in (De Smet et al., 2014b), the default
models Bertini & Dresner (Bertini, 1969, 1963) (Dresner, 1961), the
other using the CEM03 model (Mashnik et al., 2008, 1998). Fig. 2. Schematic representation of the US nozzle (manufacturer: IBA). In this exper-
iment, however, the scatterer, the range modulator and the X-/Y- collimator jaws were
not used (in order to avoid producing secondary neutrons inside the nozzle).
2. Methods and procedures

2.1. Overview of the experimental set-up

The proton therapy facility of Essen possesses a ©C230


isochronous cyclotron (manufacturer: IBA), which accelerates
protons up to a fixed energy of approximately 230 MeV. The Energy
Selection System in the cyclotron room allows reducing the proton
energy to a value in the range of 70 MeVe230 MeV according to the
specific treatment needs. The facility possesses three gantry
treatment rooms, as well as a Fixed-Beam Treatment Room (FBTR)
equipped with an eye treatment line and a multi-purpose fixed-
beam line (for e.g. intracranial, head-and-neck and prostate treat-
ments). In this experiment, the multi-purpose fixed-beam line of
the FBTR was used.
A 64  64  45 cm3 water phantom was irradiated by a
226.7 MeV proton beam delivered in Uniform Scanning (US), an
active beam delivery mode in which the beam spot is scanned in a
regular 2D pattern over a predefined area (here: 19 cm  22 cm, at
the entrance of the phantom). The shape of the proton field is
shown in Fig. 1, which is the scan of a GafChromic sheet that was
temporarily placed on the vertical entrance plane of the phantom. A
schematic representation of the US treatment nozzle is also given in
Fig. 2. In this case, however, the scatterer, the range modulator and Fig. 3. Neutron measurement positions inside and around the FBTR.

the collimators were not used and no compensator was mounted


on the snout. The beam was thus simply scanned by means of the
treatment workload was delivered during this single experiment.
scanning magnets. In this way, the production of neutrons inside
Due to this, it was necessary to scan the beam into a large field. If a
the treatment nozzle could be avoided, so that the secondary
steady beam spot directed at the isocentre had been used, there
neutron production in the FBTR arose from the water phantom
would have been a high risk of destroying the ‘IC2-3’ ionization
only. The phantom was positioned in such a way that the beam
chamber located in the nozzle (manufacturer: IBA). This chamber
isocentre lied in its vertical entrance plane, which was perpendic-
only has an average lifetime of about 1 year in routine US treatment,
ular to the beam direction (see Fig. 3).
i.e. in conditions where the delivered proton charge is spread out
A high beam current of z25 nA, i.e. more than 5 times the
over a relatively wide area of the chamber (scanned beams). When
typical treatment current, was used for the neutron measurements
the scanning magnets are switched off, the entire workload is
acquired outside the shielding. Moreover, the total irradiation time
concentrated into a spot at the centre of the ‘IC2-3’. In this case, the
of the experiment was of approximately 2 h, which is nearly 10
proton flux at the centre of the ‘IC2-3’ would have been approxi-
times the total irradiation time of one average day of treatment. In
mately 20 times larger than in the 19  22 cm2 field (supposing a
fact, it was estimated that about 20% of the average annual
spot diameter of 2.6 cm). As a consequence, the central region of
the chamber would very likely have been damaged during the
experiment if the beam had not been scanned.
To normalize the neutron measurements for the comparison
with the simulation results, it was necessary to determine the
proton charge delivered to the water phantom. This was measured
in an independent way, by means of a Bragg Peak Chamber Type
34070 (PTW, 2008) placed at the isocentre. The Bragg Peak
Chamber is an ionization chamber of ~8 cm in diameter. Since it is
not large enough for the 19  22 cm2 field, the proton charge
delivered during the neutron measurements was indirectly moni-
tored by means of the MUs recorded with the ‘IC2-3’ in the nozzle.
Fig. 1. Shape of the US field: scan of the irradiated GafChromic sheet on the vertical
The relationship between these MUs and the number of delivered
entrance plane of the water phantom.
28 V. De Smet et al. / Radiation Measurements 99 (2017) 25e40

protons measured with the Bragg Peak Chamber was established 2.2. Rem counters
during a preliminary calibration procedure. This procedure was
conducted at a low intensity of ~2.5 nA to avoid saturating the 2.2.1. WENDI-2
Bragg Peak Chamber. First, the scanning magnets of the treatment The WENDI-2 (manufacturer: Thermo Scientific) is an extended-
nozzle were switched off in order to keep the proton beam spot range rem counter designed by Olsher et al. (2000). following the
steady within the diameter of the Bragg Peak Chamber. The beam original LINUS design (Birattari et al., 1992, 1990). It has an outer
was delivered during 1 min and the electron charge collected in the diameter of 22.86 cm and consists of a 3He proportional counter
Bragg Peak Chamber was recorded. Then, without changing the surrounded by a cylindrical polyethylene assembly with an inner
beam intensity, the scanning magnets were switched on again and tungsten shell. Compared to a conventional rem counter, such as
the beam was delivered in US mode during 1 min, so as to record e.g. the Berthold LB 6411 (cf. 2.2.2), the sensitivity to high energy
the corresponding number of MUs. neutrons (E > 20 MeV) is significantly higher thanks to inelastic
The proton charge corresponding to the collected electron (n,xn) reactions occurring in the tungsten layer. The absolute
charge in the Bragg Peak Chamber was determined as: response function of the WENDI-2 was previously calculated with
MCNPX 2.5.0 (De Smet et al., 2014a) and the experimental valida-
tion with monoenergetic neutron beams between 0.144 MeV and
19 MeV can be found in (Olsher et al., 2000). The 252Cf calibration
1 nA:h
Qp ½nA:h ¼ Np  1:602  1019 C  constant of the WENDI-2 used for these measurements is 0.317
3:6  106 C nSv/count. The relative dose response function, calculated as the
With absolute response function (De Smet et al., 2014a) multiplied by
this calibration constant and divided by the fluence-to-H*(10)
conversion coefficients (ICRP, 1996; Sannikov and Savitskaya,
  1997), is shown in Fig. 4.
w J The global uncertainties associated with the WENDI-2 mea-
Qe ½C  e C
surements (4.2e7.5%) include the statistical uncertainty on the
Np ¼  
13 number of WENDI-2 counts (0.2%e6.3%), as well as the uncertainty
Scoll MeV
cm  d ½cm  1:602  10
MeV
J
on the number of delivered protons (4.2%).

where Qe is the electron charge collected in the Bragg Peak


Chamber, w/e ¼ 34.4 J/C the mean energy expended in air per unit
liberated electron charge (ICRU, 1998), d ¼ 0.2 cm the gap thickness
of the Bragg Peak Chamber (PTW, 2008), and Scoll ¼ 0.00445 MeV/ 2.2.2. LB 6411
cm the stopping power in air of 225 MeV protons (product of the The LB 6411 probe (Berthold Technologies) consists of a poly-
mass stopping power of 3.69 MeV cm2/g (ICRU, 1998), and the air ethylene moderator sphere (25 cm in diameter) with internal Cd-
density at NTP conditions of 1.205  103 g/cm3). absorbers and perforations, which surrounds a cylindrical 3He
The obtained coefficient for the conversion of the MUs into the proportional counter (Burgkhardt et al., 1997). This monitor is
proton charge delivered to the water phantom was of designed to measure thermal to 20 MeV neutrons. It is known to
1.84  104 nA h/MU. The relative standard uncertainty on this have a strongly decreasing sensitivity to neutrons above 20 MeV.
coefficient, and thus on the delivered proton charge, was estimated The 252Cf calibration factor of 0.353 nSv/count is used to display the
to 4.2%, taking into account the following uncertainties: 3.0% on the data as ambient dose equivalent H*(10). Fig. 4 also shows the
gap of the Bragg Peak Chamber (PTW, 2008), 2.0% on the mass relative dose response function of the LB 6411 as calculated by
stopping power of the protons (ICRU, 1998), 1.4% on the air density, Burgkhardt et al. up to 20 MeV (Burgkhardt et al., 1997).
1.2% on w/e (ICRU, 1998), 1% on the electron charge collected per The global uncertainties associated with the LB 6411 measure-
minute in the Bragg Peak Chamber (repeatability test) and 0.5% on ments (4.2e4.5%) include the standard uncertainty for Poisson
the number of MUs delivered per minute (repeatability test). counting (0.1e1.5%) and the uncertainty on the number of delivered
Fig. 3 gives an overview of the neutron measurement positions protons (4.2%).
inside and around the FBTR. Two positions were located inside the
treatment room (positions 1a and 1b), at respectively ~94 and ~7
from the proton beam axis, as seen from the centre of the water
phantom. Along the access maze, five different positions were used
(positions 1c to 1 g). Three positions (2a, 2b and 2c) were located
inside the technical room, which is separated from the FBTR by a
2 m thick concrete wall. Although nobody should be present inside
the technical room during irradiations performed in the FBTR, the
neutron fields in positions 2a, 2b and 2c can be considered as
representative for real workplace locations inside the proton
therapy facility. Finally, position 2d, which is located in the adjacent
gantry treatment room, is also shielded by a 2 m thick concrete
wall.
WENDI-2 measurements (cf. 2.2.1) were performed in all eleven
positions indicated in Fig. 3. The neutron spectrum was measured
with the BSS (cf. 2.3) in positions 1a, 1b, 1 g and 2a. The LB 6411 (cf.
2.2.2) was used in positions 1a, 1b, 2a, 2b and 2c. TEPC measure-
ments (cf. 2.4) were carried out in positions 1a, 1b and 2b. The
thermal neutron flux was measured with the 3He counter (cf. 2.5) Fig. 4. Relative dose response function of the WENDI-2 (De Smet et al., 2014a), the LB
in positions 1a, 1b, 2a and 2b. 6411 (Burgkhardt et al., 1997) and the TEPC (Thomas, 2014).
V. De Smet et al. / Radiation Measurements 99 (2017) 25e40 29

2.3. BSS of protons impinging on the phantom, was obtained by folding the
measured neutron spectra with the ICRP fluence-to-H*(10) con-
The CERN BSS (Agosteo et al., 2012; Birattari et al., 2010) is version coefficients (ICRP, 1996; Sannikov and Savitskaya, 1997).
composed of seven spheres: five polyethylene spheres with outer The uncertainty bars associated to the spectra unfolded with
diameters of 81, 108, 133, 178, 233 mm and two additional spheres, MAXED were calculated using the IQU code (UMG package)
nicknamed Stanlio and Ollio, containing a lead shell in order to (Reginatto et al., 2002) (Reginatto, 2004). The sensitivity analysis
extend the response up to several hundred MeV. The thermal and uncertainty propagation calculation was based on the statis-
neutron detector placed at the centre of each sphere is a 2 atm tical counting uncertainties (1%), the uncertainty on the number of
(202.65 kPa) spherical 3He proportional counter. delivered protons (4.2%) and the uncertainty on the response ma-
Monte Carlo simulations were performed with the 2011.2b trix (3%). The IQU code cannot be used for the unfolding analysis
version of the FLUKA code (Bo € hlen et al., 2014; Ferrari et al., 2005) with GRAVEL (Reginatto, 2004). However, because of the similar-
in order to recalculate, by following the same methodology, the BSS ities between the unfolding algorithms, the same uncertainties are
response matrix calculated several years ago with the 1998 version assumed here for the MAXED and GRAVEL spectra.
of FLUKA (Birattari et al., 2010). The new simulations confirmed the Since the BSS response matrix has not yet been validated in
old response matrix, improving its energy resolution (due to the quasi-monoenergetic neutron beams of high energy, the un-
new 260 group structure of the neutron cross-sections below certainties associated to the neutron fluxes above 20 MeV are
20 MeV) and reducing significantly the statistical uncertainties on assumed to be larger than those calculated using the IQU code,
the response of the two largest spheres, labelled as 233 mm and which does not deal in detail with the uncertainties associated to
Ollio (from maximum 10e20% (Birattari et al., 2010) to < 5% in this the response functions of the spheres. According to two studies
work). The new response matrix is shown in Fig. 5. The experi- about the influence of the BSS response functions on the unfolded
mental validation of the BSS response functions between 144 keV spectra (Pioch et al., 2010; Rühm et al., 2014), the neutron flux
and 19 MeV was performed at PTB in March 2002 (Birattari et al., above 20 MeV measured by a BSS outside the shielding of high-
2010). An upgraded version of the BSS for pulsed neutron fields energy particle accelerators could show variations up to 20%,
has recently been built and tested (Aza et al., 2015). depending on the Monte Carlo code and the selected in-built
The neutron spectrum is derived by unfolding the measured physics models used for the calculation of the response functions.
number of counts, normalised to the number of protons impinging This could result in variations up to 10% in the associated H*(10)
on the phantom, with the BSS response matrix by employing an rate (Pioch et al., 2010; Rühm et al., 2014). Taking also this aspect
unfolding code (Matzke, 2003). Two numerical unfolding codes into account, the global uncertainties associated to the total H*(10)
were employed: MAXimum Entropy Deconvolution (MAXED) rates measured with the BSS are evaluated at 12%.
(Reginatto and Goldhagen, 1999) and GRAVEL (Matzke, 1994). The
first is a code specially designed for multi-sphere neutron spec-
trometer data, whereas the second has more general applications. 2.4. TEPC
Due to the nonuniqueness of the unfolding process, physically
meaningful hypotheses must be formulated to counterbalance the The TEPC (Model LET-SW5, Far West Technology) used for these
lack of information. Consequently, these two codes need an a priori measurements is spherical, with an internal diameter of 12.55 cm
estimation of the true spectrum, called guess spectrum. This was and a 2 mm thick shell of A-150 tissue-equivalent plastic. The TEPC
obtained via Monte Carlo simulations (see section 2.6). MAXED and was filled with propane tissue-equivalent gas at a pressure of 8.8
GRAVEL iteratively alter this guess spectrum on the basis of mBar to simulate a tissue site size of 2 mm (Waker, 1995). The
different algorithms by attempting to reach the best agreement relative dose response function of this type of detector, calculated
between the measured and the estimated counts, obtained by up to 20 MeV by D.J. Thomas (2014), is shown in Fig. 4.
folding the response matrix with the guess spectrum. Microdosimetric dose distributions, as a function of equal log-
The ambient dose equivalent, H*(10), normalised to the number arithmic intervals of the lineal energy y (keV/mm), were constructed
by applying the principles explained in (Waker, 1995). The

Fig. 5. BSS response functions, obtained via FLUKA simulations.


30 V. De Smet et al. / Radiation Measurements 99 (2017) 25e40

calibration of the microdosimetric spectra was based on the proton (McConn et al., 2011), detailed in Table 1, was defined with a density
edge, which is a distinctive feature in the spectra defined as the of 2.30 g/cm3.
maximum lineal energy imparted over the full chord length. For a To obtain sufficient statistics for positions located outside the
2 mm simulated site size, this quantity equals 136 keV/mm (Waker, FBTR or at the end of the access maze, the variance reduction
1995). In the proton edge region, the microdosimetric spectra technique named “geometry splitting and Russian roulette” (X-5
were fitted with a Fermi-like function and the value of 136 keV/mm Monte Carlo Team, 2003) was used. By assigning an appropriate
was assigned to the intercept of the tangent at the inflexion point of neutron importance to each cell, the neutron populations that
this function (Conte et al., 2013). The measured lineal energy range migrated from the source region towards the tally regions were
spans from 2 keV/mm to 230 keV/mm. maintained approximately constant. MCNPX adjusted the particle
The absorbed dose distribution was obtained with the following weights accordingly to this, to avoid biasing the tally results. In our
formula (Waker and Aslam, 2010): simulated geometry, the concrete shielding walls were divided in
  several cells of 20 cm in thickness, perpendicularly to the main
yi dðyi Þ ½keV=mm  l ½mm 16 J direction of the neutron flux. The chosen cell thickness allowed
Di ðGyÞ ¼       1:602  10
r kg m3  V m3 keV keeping all importance ratios of adjacent concrete cells below a
factor of 8, as recommended for this variance reduction technique
where the mean chord length l ¼ 1.333 mm, the gas density (Booth, 1985). Geometry splitting was also applied to the air vol-
r ¼ 0.0157 kg/m3 and the gas volume V ¼ 0.001073 m3. umes inside the access maze, perpendicularly to the main direction
The dose equivalent H was calculated as the integral of the of the neutron flux along the maze. The distribution of neutron
product of the absorbed dose distribution with the quality factor importances is represented graphically in Fig. 6. A total of
function Q (ICRU, 1980). 1.875  1011 histories were simulated. The simulation was sub-
The uncertainties on H include the standard deviation of three divided in 25 runs of different pseudorandom number sequences,
consecutive measurements (normalised per MU) acquired in the which ran on 25 processors during 2.5 weeks. The tally results xi of
same position (1.4%e2.4%) and the uncertainty on the number of the 25 runs were then combined by computing their average x
protons delivered per MU (4.2%). weighted by the corresponding MCNPX absolute uncertainties si :
P vffiffiffiffiffiffiffiffiffiffi
2.5. 3
He-filled proportional counter
1
i s2 xi
u
u 1
x¼ P i
1
and sx ¼ tP 1
i s2 i s2
The proportional counter type 70061 (manufacturer: VacuTec) i i

consists of a stainless steel tube (type AISI 316L) filled with 3He gas,
The available endf70prot (Trellue et al., 2008) and endf70
with an active volume of 6.2 cm3. The thermal neutron sensitivity is
(Trellue et al., 2009) cross-section libraries, based on the ENDF/B-
3.6 counts/(n/cm2). The device is operating at 960 V and
VII.0 evaluation (Chadwick et al., 2006), were used for the in-
8 bar ± 15%. The detector response strongly decreases for neutrons
teractions of protons and neutrons below 150 MeV. Above 150 MeV,
with energies higher than thermal.
physics models have to be selected within the code in order to
The calibration of the 3He counter was performed with a colli-
calculate the intranuclear cascade (INC) and evaporation processes
mated thermal neutron beam (Belgian Reactor 1, BR1 at SCK-CEN).
of the hadron interactions (Pelowitz, 2011). For the positions inside
A correction is performed for non-collimated fields through the
the technical room, the default Bertini & Dresner models (for the
calculation of the mean angular response of the counter. Finally
these coefficients are used together with the fluence-to-H*(10)
conversion coefficient (10.6 pSv.cm2 (ICRP, 1996)) to determine
the ambient dose equivalent H*(10) for thermal neutrons.
The global uncertainties associated with the thermal H*(10)
measurements were estimated at 26%, which includes un-
certainties related to the Poisson counting (0.08e1.2%), the number
of delivered protons (4.2%), variations in the calibration process (1%,
mainly due to the uncertainties on the neutron flux and the reactor
power), and corrections for the angular response (25%).

2.6. MCNPX simulation of the treatment room

The full FBTR was simulated using MCNPX 2.7.0 (Pelowitz, 2011).
The particle source was defined in front of the water phantom as a
19 cm  22 cm rectangle producing a parallel and uniform beam of
226.7 MeV protons. It was assumed that, with regards to the
simulation of the secondary neutron production in the FBTR, this
source definition provides a sufficiently good approximation of the
true proton field shown in Fig. 1. The real concrete composition of
the FBTR being unknown, the standard ‘NIST Portland’ composition

Table 1
Elemental composition (% atoms) of ‘NIST Portland’ concrete (2.30 g/cm3) (McConn
et al., 2011).
Fig. 6. Distribution of neutron importances for the variance reduction in the FBTR
H C O Na Mg Al Si K Ca Fe simulation. The smallest importance ¼ 1 (dark blue cells) and the largest
importance ¼ 1024 (red cells). (For interpretation of the references to colour in this
16.88 0.14 56.25 1.18 0.14 2.14 20.41 0.56 1.87 0.43
figure legend, the reader is referred to the web version of this article.)
V. De Smet et al. / Radiation Measurements 99 (2017) 25e40 31

INC and the evaporation, respectively) were identified as yielding simulated WENDI-2 response in a given measurement position was
the highest H*(10) values, whereas the CEM03 model (computing calculated as the average Rsim,avg of the five energy-integrated re-
both the INC and the evaporation) was found to be the least con- sponses Rsim,i obtained with the abovementioned response func-
servative (De Smet, 2016). Therefore, two versions of the FBTR tions. For the standard uncertainty on Rsim,avg, it was assumed that
simulation are compared in this paper: one using Bertini & Dresner the used response functions represent a coverage factor of 3. A
and the other using CEM03. The H*(10) rates as well as the WENDI- symmetric uncertainty sresp.fct., representing the uncertainties on
2 responses were calculated in both simulations. the response function below 0.1 MeV and above 20 MeV, was thus
estimated as:
2.6.1. Simulation of the H*(10) rates
1  
Neutron spectra were calculated inside air spheres of 25 cm in sresp:fct: ¼ max R  Rsim;avg 
diameter located in the eleven reference positions indicated in 3 1i5 sim;i
Fig. 3. The associated H*(10) rates were computed using the
The total uncertainty stot on Rsim,avg, which combines the
fluence-to-H*(10) conversion coefficients from ICRP74 (ICRP, 1996)
average statistical MCNPX uncertainty sstat and the uncertainties on
and its high energy extension from Sannikov & Savitskaya
the entire WENDI-2 response function, was calculated as:
(Sannikov and Savitskaya, 1997). The uncertainties associated to the
H*(10) results (see sections 3.3 and 3.5) are purely statistical. qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
stot ¼ s2stat þ s2exp þ s2resp:fct:
2.6.2. Simulation of the WENDI-2 responses
The simulated neutron spectra were also used to calculate the
response of the WENDI-2 at the eleven reference positions. In order
to simulate the WENDI-2 response, one has to fold the simulated 3. Discussion of the results
neutron spectrum with the response function of the detector.
In the range of 0.1 MeVe20 MeV, the WENDI-2 response func- 3.1. Simulated spectra
tion has already been successfully characterised by means of e.g.
monoenergetic neutron beam measurements (Olsher et al., 2000). For the positions 1a, 1b, 2a and 1 g, Figs. 8e12 show the simu-
Outside of this energy range, however, the response function is less lated neutron spectra obtained with the Bertini & Dresner and the
well characterised. Several versions of the WENDI-2 response CEM03 models, respectively. These figures also show the BSS
function, calculated by different authors with different Monte Carlo unfolding results, which will be discussed in section 3.2.
codes, exist in literature (De Smet, 2016; De Smet et al., 2014a; Position 1b is located close to the proton beam axis inside the
€gerhofer et al., 2012; Olsher et al., 2000; Vanaudenhove et al.,
Ja treatment room. The neutron spectrum in this position is charac-
2014) and show differences up to a factor of 1.5 in the range of terised by a large INC peak (10 MeVe230 MeV) because the emis-
20 MeVe230 MeV. Unfortunately, validating the response function sion of high-energy neutrons during INCs is strongly forward-
experimentally at high energies is very challenging. Truly mono- peaked (see Fig. 9). In comparison, the INC peak in position 1a,
energetic neutron beams cannot be produced because the high located at approximately 90 from the proton beam axis, is indeed
energy of the charged projectile exceeds the threshold for multi- much smaller (see Fig. 8).
body break-up reactions. This creates a continuous low-energy In position 2a (see Fig. 11), the simulated spectrum exhibits the
tail in the neutron spectrum, which is referred to as “quasi-mon- typical shape of neutron fields outside the shielding of proton ac-
oenergetic”. The presence of this low-energy tail constitutes a celerators: it is characterised by an INC peak around ~100 MeV, an
complication because extended-range rem meters like the WENDI- evaporation peak around ~2 MeV and a thermal peak (around 5
2 have a good sensitivity to both the peak and tail neutrons. In fact, x10-7 MeV). The INC peak represents high-energy neutrons pro-
no international standard exists for this type of calibration at high duced during the pre-equilibrium stage of non-elastic hadron in-
energies (Harano and Nolte, 2011). The literature contains very few teractions (INCs). Most of these neutrons have undergone very few
experimental validation data for the WENDI-2 response function interactions and travelled in nearly straight line through the shield.
above 20 MeV. Olsher & McLean published a few values based on The evaporation peak essentially contains neutrons which were
quasi-monoenergetic neutron beam measurements with peak en- produced in the outer part of the shield, during the equilibrium
ergies in the range of 45 MeVe175 MeV (Olsher and McLean, 2008). stage of non-elastic hadron interactions (evaporation). The thermal
As shown in Fig. 7, these experimental values are lower than the peak corresponds mainly to room-scattered neutrons.
simulated WENDI-2 response functions by a factor of 2 on average. The spectrum of position 1g at the end of the maze (see Fig. 12)
Recently, a similar experiment with high-energy quasi-mono- has a large thermal peak due to neutrons that have moved along the
energetic neutron beams (De Smet, 2016) yielded values that are maze path and have lost their kinetic energy progressively through
intermediate to e and statistically consistent with e the GEANT4 many scattering interactions. The spectrum also possesses a small
BIC response function (Vanaudenhove, 2014) and the first experi- INC and evaporation peak from neutrons that have been directly
mental data (Olsher and McLean, 2008). transmitted through the walls of the technical room.
In order to assess the uncertainties on the (energy-integrated) As shown in Figs. 8e12, the neutron spectra simulated with the
WENDI-2 responses obtained in the simulations of the FBTR, it was Bertini & Dresner models are larger than those simulated with the
verified how much these responses could vary according to the CEM03 model, on the entire energy range. For positions inside the
selected version of the WENDI-2 response function. Each neutron treatment room (1a and 1b), the differences are more pronounced
spectrum was therefore folded with the five versions of the in the INC peak (10 MeVe230 MeV) than at lower energies. In
response function listed in Table 2 and shown in Fig. 7. In the range position 1b, they correspond to factors of 1.2e2.0 above 10 MeV and
of 0.1 MeVe20 MeV, where the response function is considered ~1.3 below 10 MeV (see Fig. 9). For position 1a, this trend is not
well characterised, a common section for all response functions clearly visible in Fig. 8 because the fraction of neutrons above
was defined based on the monoenergetic irradiation data (Olsher ~100 MeV is very small. There are nevertheless differences by fac-
et al., 2000). A relative standard uncertainty of 5%, corresponding tors of 1.2e2.6 above 10 MeV and ~1.2 below 10 MeV.
to the experimental uncertainty in (Olsher et al., 2000), was taken Due to the propagation mechanisms of the INC through the
into account in this energy range (sexp). The reference value for the concrete shield, a lot of additional neutrons are produced at all
32 V. De Smet et al. / Radiation Measurements 99 (2017) 25e40

Fig. 7. Five versions of the dose response function used to simulate the WENDI-2 response in the measurement positions around the FBTR (De Smet, 2016; Ja€gerhofer et al., 2012;
Olsher et al., 2000; Olsher and McLean, 2008; Vanaudenhove, 2014), based on a calibration constant of 0.317 nSv/count. A common section for all response functions was defined
between 0.144 MeV and 19 MeV, based on experimental data from monoenergetic neutron beam irradiations (Olsher et al., 2000). The fluence-to-H*(10) conversion coefficients
(ICRP, 1996) are also shown in this graph.

Table 2
Versions of the simulated WENDI-2 response function used in this work. (N.B.: the version simulated with MCNP4B only reaches up to 20 MeV. In this work, it was coupled to
the experimental data at high energies of Olsher and McLean, see Fig. 7).

Reference Monte Carlo code High-energy physics models

(Olsher et al., 2000) MCNP4B (Briesmeister, 1997) use of evaluated cross-section data only, because energies lower than 20 MeV
€gerhofer et al., 2012)
(Ja FLUKA (Ferrari et al., 2005) FLUKA GINC & evaporation (Ferrari et al., 2005), above 20 MeV
(Vanaudenhove, 2014) GEANT4 9.6 (Agostinelli et al., 2003; Allison et al., 2006) Bertini INC (Bertini, 1969, 1963) & Weisskopf (Weisskopf, 1937), above 20 MeV
(Vanaudenhove, 2014) GEANT4 9.6 (Agostinelli et al., 2003; Allison et al., 2006) BIC INC (Folger et al., 2004) & Weisskopf (Weisskopf, 1937), above 20 MeV
(De Smet, 2016) MCNPX 2.7.0 (Pelowitz, 2011) Bertini (Bertini, 1969, 1963) & Dresner (Dresner, 1961), above 150 MeV

depths and the initial surplus of high-energy neutrons in the Bertini For all positions, the spectra unfolded with GRAVEL matched
& Dresner spectrum in front of the shielding eventually results in with the MAXED results within 1s uncertainties. For the sake of
an overestimation factor that affects the whole energy range rela- readability, only the MAXED spectra are shown in Figs. 8e12.
tively uniformly, behind the shielding. In position 2a, for example,
both simulated spectra have very similar shapes, the Bertini &
Dresner spectrum being larger than the CEM03 spectrum by a 3.2.1. Inside the treatment room
factor of ~1.6e1.7 at all energies (see Fig. 11). Position 1 g is only In Position 1a (Fig. 8), situated at roughly 90 from the proton
partially affected by this effect (see Fig. 12) because the main flux beam axis, the guess spectrum e simulated with the Bertini &
originates from neutron scattering inside the access maze rather Dresner models e matches the unfolded spectra very well for
than from the direct transmission of neutrons through the walls of neutrons with energies larger than ~5 MeV. In the intermediate
the technical room. energy range and in the evaporation peak up to ~5 MeV, the guess
spectrum is smaller than the unfolded spectra by ~20%. The
agreement is better than for the spectrum simulated with the
3.2. Bonner Sphere Spectrometry vs. simulated spectra CEM03 model, which underestimates the unfolded spectra by ~30%
in the intermediate, evaporation and high-energy regions.
The neutron spectra calculated with the Bertini & Dresner For Position 1b, located in the direction of forward neutron
models were used as guess spectra for unfolding the BSS data. For emission, the simulated and measured spectra are compared in
position 1b, where the shape of the simulated spectrum depends Fig. 9 and a zoom on the INC peak is shown in Fig. 10. For the sake of
more strongly on the selected physics models, the unfolding was readability, the uncertainty bars of the two spectra unfolded with
also performed using the CEM03 spectrum as a guess. MAXED are replaced in the plots by two dotted curves that delimit
In Figs. 8e12, the uncertainties associated to the spectra their envelope. The comparison of the unfolding results obtained
unfolded with MAXED are those calculated with the IQU code with the two different guess spectra bring out the fact that, as
(Reginatto, 2004), except for energies above 20 MeV where 20% predicted by (Reginatto, 2009), the energy resolution of the BSS at
relative uncertainties were considered (see section 2.3). high energies is not sufficient to determine the exact shape of the
V. De Smet et al. / Radiation Measurements 99 (2017) 25e40 33

Fig. 8. Neutron spectrum in position 1a, as obtained by the MCNPX simulations using the Bertini & Dresner models (in black) and the CEM03 model (in purple), and as obtained by
unfolding of the BSS data using MAXED with the MCNPX Bertini & Dresner spectrum as a guess (in green). The statistical uncertainties on the simulated spectra are 1% for bins up
to 70 MeV and 10% for bins above 70 MeV. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 9. Neutron spectrum in position 1b, as obtained by the MCNPX simulations using the Bertini & Dresner models (in black) and the CEM03 model (in purple), compared to the
unfolded BSS spectra obtained using MAXED with as a guess (a) the MCNPX Bertini & Dresner spectrum (in green), and (b) the MCNPX CEM03 spectrum (in yellow). In grey dotted
lines: the envelope of the uncertainties associated to the two BSS spectra unfolded with MAXED. Uncertainties on the simulated spectra are 1% for all bins. (For interpretation of
the references to colour in this figure legend, the reader is referred to the web version of this article.)

INC peak without ambiguity. Nevertheless, it is interesting to note (Satoh et al., 2012).
that the spectrum simulated with the Bertini & Dresner models At the end of the access maze, in position 1 g, the neutron
agrees very well with the unfolded spectra at all energies except in spectrum (Fig. 12) is mainly characterszed by a large thermal peak
the INC peak, where an overestimate appears above 100 MeV. The as well as a relatively small INC peak. The simulated spectra are
spectrum simulated with the CEM03 model, however, only larger than the average BSS results on the whole energy range, but
matches with the unfolding results in the INC peak and un- the overestimation factor is actually larger on the INC peak
derestimates them at all other energies. (contribution from neutrons directly transmitted through the
technical room) than on the thermal peak (main contribution from
3.2.2. Outside the treatment room neutrons scattering inside the access maze).
In Position 2a, located at roughly 40 from the proton beam axis,
the unfolded spectra (Fig. 11) are lower than the simulated spectra 3.3. Simulated WENDI-2 responses and H*(10)
by a factor that remains relatively constant on the entire energy
range (a factor ~2.3 w.r.t. the Bertini & Dresner spectrum and ~1.4 The energy-integrated WENDI-2 responses simulated using the
w.r.t. the CEM03 spectrum). In this aspect, the results are similar to response functions of Fig. 7 are given in Table 3 for the simulation of
those obtained by Satoh et al. with the simulation code PHITS and the FBTR based on the Bertini & Dresner models. The same results
the DARWIN spectrometer at the Fukui Proton Therapy Centre for the FBTR simulation based on the CEM03 model are given in
34 V. De Smet et al. / Radiation Measurements 99 (2017) 25e40

Table 4. The average simulated WENDI-2 responses are then pro-


vided in Table 5, for both FBTR simulations. As shown in this table,
the uncertainties on the detector response function lead to stan-
dard uncertainties of 4%e10% on the average simulated WENDI-2
responses.
The simulated H*(10) obtained in the two FBTR simulations are
given in Table 6. In both simulations, the average simulated WENDI-
2 responses agree with the simulated H*(10) within 1s un-
certainties in the positions 2ae2d outside the treatment room, and
the positions 1de1g in the maze. The theoretical accuracy of the
WENDI-2 varies between 3% and þ8% in these positions. Inside
the treatment room (1a, 1b) and at the entrance of the maze (1c),
the average simulated WENDI-2 responses mostly agree with the
simulated H*(10) within 2s uncertainties. The theoretical accuracy
lies between þ7% and þ13% for position 1b and between 7%
and 9% for positions 1a and 1c. According to these simulations, the
WENDI-2 should thus provide relatively accurate neutron H*(10)
Fig. 10. Zoom on the INC peak of the spectra shown in Fig. 9. measurements inside and around the FBTR.

Fig. 11. Neutron spectrum in position 2a, as obtained by the MCNPX simulations, using the Bertini & Dresner models (in black) and the CEM03 model (in purple), and as obtained by
unfolding of the BSS data using MAXED with the MCNPX Bertini & Dresner spectrum as a guess (in green). The statistical uncertainties on the simulated spectra are 3% for all bins.
(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 12. Neutron spectrum in position 1 g, as obtained by the MCNPX simulations using the Bertini & Dresner models (in black) and the CEM03 model (in purple), and as obtained by
unfolding of the BSS data using MAXED with the MCNPX Bertini & Dresner spectrum as a guess (in green). The statistical uncertainties on the simulated spectra are 3% for all bins.
(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
V. De Smet et al. / Radiation Measurements 99 (2017) 25e40 35

Table 3
Simulation of the FBTR based on the Bertini & Dresner models: simulated WENDI-2 responses obtained using the different versions of the response function shown in Fig. 7.
The results are expressed in mSv/(nA.h), where the nA.h refer to the proton charge delivered to the phantom. The given uncertainties are the MCNPX statistical uncertainties
(1s).

Position MCNPX 2.7.0 Bertini & Dresner GEANT4 9.6 Bertini GEANT4 9.6 BIC FLUKA MCNP4B þ exp. data at high energies

1b 2460 ± 1 2758 ± 2 2289 ± 1 2385 ± 1 1667 ± 1


1a 530.8 ± 0.5 555.0 ± 0.5 537.4 ± 0.5 521.8 ± 0.5 510.8 ± 0.5
1c 373.3 ± 0.4 382.7 ± 0.5 354.7 ± 0.4 358.3 ± 0.5 312.2 ± 0.4
1d 33.2 ± 0.1 35.9 ± 0.1 35.2 ± 0.1 32.1 ± 0.1 33.9 ± 0.1
1e 10.28 ± 0.05 11.28 ± 0.05 11.11 ± 0.05 9.75 ± 0.05 10.71 ± 0.05
1f 3.88 ± 0.02 4.37 ± 0.02 3.98 ± 0.02 3.67 ± 0.02 3.38 ± 0.02
1g 3.25 ± 0.02 3.68 ± 0.02 3.15 ± 0.02 3.11 ± 0.02 2.41 ± 0.02
2a 6.67 ± 0.03 7.35 ± 0.03 6.44 ± 0.03 6.59 ± 0.02 5.14 ± 0.02
2c 17.10 ± 0.04 18.99 ± 0.04 16.14 ± 0.04 16.83 ± 0.04 12.24 ± 0.03
2b 12.13 ± 0.03 13.41 ± 0.03 11.55 ± 0.03 11.93 ± 0.03 8.98 ± 0.03
2d 0.099 ± 0.001 0.108 ± 0.001 0.098 ± 0.001 0.098 ± 0.001 0.082 ± 0.001

Table 4
Simulation of the FBTR based on the CEM03 model: simulated WENDI-2 responses obtained using the different versions of the response function shown in Fig. 7. The results are
expressed in mSv/(nA.h), where the nA.h refer to the proton charge delivered to the phantom. The given uncertainties are the MCNPX statistical uncertainties (1s).

Position MCNPX 2.7.0 Bertini & Dresner GEANT4 9.6 Bertini GEANT4 9.6 BIC FLUKA MCNP4B þ exp. data at high energies

1b 1738 ± 1 1913 ± 1 1616 ± 1 1687 ± 1 1211 ± 1


1a 450.5 ± 0.5 459.6 ± 0.5 447.5 ± 0.5 431.3 ± 0.5 429.2 ± 0.5
1c 289.2 ± 0.4 300.5 ± 0.4 281.1 ± 0.4 281.7 ± 0.4 251.5 ± 0.4
1d 28.29 ± 0.09 29.14 ± 0.09 28.65 ± 0.09 25.92 ± 0.09 27.69 ± 0.09
1e 8.42 ± 0.04 8.76 ± 0.04 8.64 ± 0.04 7.49 ± 0.04 8.35 ± 0.04
1f 2.72 ± 0.02 2.96 ± 0.02 2.72 ± 0.02 2.44 ± 0.02 2.33 ± 0.02
1g 2.05 ± 0.02 2.28 ± 0.02 1.96 ± 0.02 1.91 ± 0.02 1.51 ± 0.01
2a 4.06 ± 0.02 4.40 ± 0.02 3.86 ± 0.02 3.95 ± 0.02 3.07 ± 0.02
2c 10.28 ± 0.03 11.31 ± 0.03 9.61 ± 0.03 10.04 ± 0.03 7.26 ± 0.02
2b 6.66 ± 0.02 7.27 ± 0.03 6.27 ± 0.02 6.49 ± 0.02 4.89 ± 0.02
2d 0.0589 ± 0.0006 0.0625 ± 0.0006 0.0567 ± 0.0006 0.0570 ± 0.0005 0.0479 ± 0.0005

Table 5
Average simulated WENDI-2 responses obtained in the FBTR simulation based on the Bertini & Dresner models and in the FBTR simulation based on the CEM03 model. The
results are expressed in mSv/(nA.h). The uncertainties (1s) were calculated as explained in section 2.6.2.

Position Spectra: Bertini & Dresner Spectra: CEM03

Rsim,avg relative uncertainty Rsim,avg relative uncertainty

1b (2.3 ± 0.2)  103 9% (1.6 ± 0.2)  103 10%


1a (5.3 ± 0.2)  102 4% (4.4 ± 0.2)  102 5%
1c (3.6 ± 0.2)  102 5% (2.8 ± 0.2)  102 6%
1d (3.4 ± 0.1)  101 4% (2.8 ± 0.2)  101 6%
1e (1.06 ± 0.05)  101 4% (8.3 ± 0.5)  100 6%
1f (3.9 ± 0.2)  100 5% (2.6 ± 0.2)  100 6%
1g (3.1 ± 0.2)  100 8% (1.9 ± 0.2)  100 9%
2a (6.4 ± 0.5)  100 7% (3.9 ± 0.3)  100 9%
2c (1.6 ± 0.1)  101 8% (9.7 ± 0.9)  100 10%
2b (1.16 ± 0.09)  101 8% (6.3 ± 0.6)  100 9%
2d (9.7 ± 0.6)  10-2 6% (5.7 ± 0.4)  10-2 7%

3.4. WENDI-2 measurements vs. simulation results maximum above 100 MeV, the average simulated WENDI-2
response based on the Bertini & Dresner spectrum seems to
The WENDI-2 measurements are given in Table 7. The ratios of slightly overestimate the WENDI-2 measurement. This might
the simulation results to the WENDI-2 measurements are shown in reflect the overestimate on the high-energy peak observed with
Fig. 13. respect to the BSS results above 100 MeV (see Fig. 10).

3.4.1. Inside the treatment room 3.4.2. Outside the treatment room
Inside the treatment room (1a, 1b) and the first part of the maze In the adjacent rooms (2a, 2b, 2c, 2d) and at the end of the maze
(1c, 1d, 1e), the simulated WENDI-2 responses based on the Bertini (1 g), the simulation results obtained with the Bertini & Dresner
& Dresner spectra agree with the WENDI-2 measurements within models and the CEM03 model both overestimate the WENDI-2
one to two standard uncertainties (see Fig. 13). The agreement is measurements (see Fig. 13). The overestimate ranges up to a fac-
better than for the simulation results based on the CEM03 model, tor of ~3 with the Bertini & Dresner models, and up to a factor of
which underestimate the measurements by a factor of 1.2e1.3. This ~1.7 with the CEM03 model. In both cases, the simulations are thus
is globally consistent with our observations on the BSS spectra (see conservative with respect to the wall-transmitted neutron fluxes.
section 3.2.1). Such conservative behaviour has also been reported in (De Smet
In position 1b, characterised by a large high-energy peak with a et al., 2014b; Satoh et al., 2012), with overestimates by factors of
36 V. De Smet et al. / Radiation Measurements 99 (2017) 25e40

Table 6 outside the shielding (De Smet, 2016). In this case, however, such
Simulated H*(10) obtained in the FBTR simulation based on the Bertini & Dresner uncertainties do probably not play a significant role. There were
models and in the FBTR simulation based on the CEM03 model. The results are
expressed in mSv/(nA.h). The uncertainties (1s) are the statistical MCNPX
indeed no equipment parts or cupboards located between the
uncertainties. water phantom and the wall of the technical room (see positions
2ae2c). Also, the high-energy neutron flux emitted towards posi-
Position Spectra: Bertini & Dresner Spectra: CEM03
tion 2d was not significantly influenced by the unmodelled eye
H*(10) H*(10) treatment line since the latter did not cross the line-of-sights from
1b (2.054 ± 0.001)  103 (1.526 ± 0.009)  103 the phantom to 2d.
1a (5.782 ± 0.006)  102 (4.773 ± 0.005)  102 As for the variance reduction technique necessary to achieve the
1c (3.889 ± 0.005)  102 (3.088 ± 0.004)  102
statistical consistency for the wall-transmitted fluxes, we assume
1d (3.51 ± 0.01)  101 (2.83 ± 0.01)  101
1e (1.078 ± 0.005)  101 (8.33 ± 0.04)  100 that it must not have introduced a significant bias in the simulation
1f (3.74 ± 0.02)  100 (2.54 ± 0.02)  100 results since it was applied according to the general recommen-
1g (2.89 ± 0.02)  100 (1.81 ± 0.02)  100 dations of the field.
2a (6.28 ± 0.03)  100 (3.78 ± 0.02)  100 Uncertainties related to the material definition of the concrete
2c (1.52 ± 0.04)  101 (9.13 ± 0.03)  100
might, on the other hand, be significant. The true elemental
2b (1.10 ± 0.03)  101 (6.03 ± 0.03)  100
2d (9.9 ± 0.1)  10-2 (5.75 ± 0.05)  10-2 composition is indeed unknown and the density is expected to be
slightly underestimated in the simulations. A recent sensitivity
study (De Smet, 2016) reported possible overestimates by up to
Table 7
factors of 1.3e1.4 at forward angles and 1.5e1.6 at lateral angles for
WENDI-2 measurements, expressed in (mSv/(nA.h)). the H*(10) outside the shielding, when comparing the reference
composition ‘NIST Portland’ to twelve other ordinary concrete
Position WENDI-2 measurements
compositions of greater shielding effectiveness. According to this
1b (2.03 ± 0.09)  103 study, the concrete composition and density can indeed lead to
1a (5.8 ± 0.2)  102
1c (3.8 ± 0.2)  102
non-negligible errors on the simulated neutron H*(10) outside the
1d (3.6 ± 0.2)  101 shielding, but they can probably not be the single cause of the
1e (1.04 ± 0.05)  101 observed simulation-to-measurement ratios.
1f (2.8 ± 0.2)  100 Uncertainties related the traversed concrete thickness may also
1g (1.4 ± 0.1)  100
play a role, since in practice a constructed building is never the
2b (2.7 ± 0.1)  100
2c (5.6 ± 0.2)  100 perfect reflection of its architectural plans. These uncertainties may
2a (3.9 ± 0.2)  100 not necessarily be large, but might be non-negligible for such a
2d (5.3 ± 0.3)  10-2 thick concrete shield (De Smet, 2016).

Fig. 13. The following quantities are given in ratio to the WENDI-2 measurements: (a) the average simulated WENDI-2 response based on the Bertini & Dresner spectra, (b) the
neutron H*(10) rates based on the Bertini & Dresner spectra, (c) the average simulated WENDI-2 response based on the CEM03 spectra, (d) the neutron H*(10) rates based on the
CEM03 spectra.

2e7 on the neutron H*(10) simulated outside the treatment rooms. Positioning uncertainties affecting the simulated H*(10) were
The overestimates are most likely due to a combination of un- only investigated in the Bertini & Dresner case, in which the tally
certainties affecting several simulation parameters. Future work spheres were systematically displaced by 10 cm in 6 different di-
should be dedicated to investigating the various sources of rections. The largest deviations obtained were of 3% for positions
uncertainties. 1ae1c; 5e6% for positions 1de1g; and 4e7% for positions 2ae2d.
Unmodelled objects causing a non-negligible attenuation of the The influence of uncertainties on the interaction cross-sections
high-energy neutron flux inside the treatment room can some- certainly need to be further investigated. The uncertainties are
times have a large impact on the accuracy of the simulated H*(10) larger for the model-based cross-sections above 150 MeV than for
V. De Smet et al. / Radiation Measurements 99 (2017) 25e40 37

the evaluated cross-sections below 150 MeV, since the latter are The microdosimetric spectra of positions 1a and 1b show good
partially based on experimental cross-section data. Furthermore, agreement with TEPC spectra acquired in a proton therapy centre
the uncertainties on the evaluated data are expected to be larger in during SOBP treatment (Pe rez-Andújar et al., 2012).
the range of 20 MeVe150 MeV than below 20 MeV, due to the The neutron absorbed dose for position 1b can be mainly
lower availability of experimental data and the often larger attributed to high-energy neutrons (recoil protons with low stop-
measuring uncertainties above 20 MeV. Future work is necessary to ping powers; see the peak around 15 keV/mm), with some contri-
determine whether the uncertainties on the cross-sections for butions from low-energy neutrons (recoil protons with high
neutron (and proton) interactions inside the concrete significantly stopping powers). In position 1a, the spectrum shows a slightly
contribute to the overestimate on the neutron attenuation length. higher contribution of low-energy neutrons (see the peak around
Special attention should also be paid to the high-energy end of the 50 keV/mm) compared to high-energy neutrons. This is consistent
double-differential neutron production cross-section for protons with the neutron spectrum measured by the BSS in position 1a (see
on oxygen nuclei (interactions in the water phantom). The flux of Fig. 8). Beyond the proton edge, the contribution of alpha particles
high-energy neutrons (E > 100 MeV) emitted by the phantom is also higher for position 1b than for position 1a. This confirms a
indeed strongly determines the neutron H*(10) outside the higher contribution of high-energy neutrons to the dose in position
shielding, as illustrated to some extent in the comparison of the 1b, as also observed in the BSS spectrum (see Fig. 9).
simulations based on the Bertini and CEM03 models (section 3.1). For position 2b, the proton edge is less pronounced while the
According to our BSS spectrum measured in position 1b, the Bertini contributions of high- and low-energy neutrons are comparable, as
model seems to slightly overestimate the forward-angle production it is also observed in the BSS spectra for the technical room (posi-
of neutrons of more than 100 MeV in the water phantom (see tion 2a, Fig. 11). The dose contribution of gamma-induced events is
Fig. 10). This aspect might thus contribute in a non-negligible way clearly larger in the technical room (position 2b) than inside the
to the overestimate observed in position 2b for the H*(10) simu- treatment room (positions 1a and 1b). As mentioned above, these
lated with the Bertini & Dresner models. events of low lineal energy (<10 keV) may also be attributed to the
contribution of thermal neutrons.
3.5. TEPC spectra
3.6. Global intercomparison of the measurements
Fig. 14 shows the microdosimetric spectra acquired in positions
1a, 1b and 2b. The overall interpretation of TEPC spectra is chal- The neutron H*(10) rates obtained from unfolding the BSS data
lenging due to the overlap between the different stopping powers with MAXED and GRAVEL are in good agreement with each other in
of the different types of particles. Lineal energies lower than all positions (see Table 8). In the case of position 1b, a good
10 keV/mm are gamma-induced events (production of recoil elec- agreement also exists between the H*(10) rates calculated using the
trons), although in this region thermal neutrons may also unfolded spectra based on the Bertini & Dresner and the CEM03
contribute to the signal (Waker, 1995). From 10 keV/mm up to the guess spectra.
proton edge (136 keV/mm for a 2 mm tissue simulated site size As shown in Table 9, the WENDI-2 measurements agree within
(Waker, 1995)), the events are attributed to recoil protons resulting ~10% with the H*(10) rates of the BSS data in the four types of
from the elastic scattering of neutrons on hydrogen nuclei. Low neutron fields (positions 1a, 1b, 1g, 2a). It corresponds to an
lineal energies in this range correspond to energetic recoil protons agreement within one standard uncertainty. There thus seems to be
(i.e. protons with low stopping powers), which are produced by little need to introduce position-specific calibration correction
high-energy neutrons. The higher lineal energies, on the contrary, factors, based on the local neutron spectrum, in order to improve
are attributed to less energetic recoil protons, produced by less the accuracy of the H*(10) measurements performed with the
energetic neutrons. Beyond the proton edge, events due to alpha WENDI-2.
particles generated in (n,a) reactions are recorded. The cross- Compared to the WENDI-2, the conventional rem meter LB 6411
section of this reaction becomes only important for high-energy measures values lower by 40% in position 1b and by 27%e28% in
neutrons. positions 2a, 2b, and 2c, which are located behind the shielding in
forward direction with respect to the impinging proton beam (see
Table 9). The low response of the LB 6411 to high-energy neutrons
thus causes a small underestimate of the measured H*(10) rates in
these positions in which the proportion of neutrons with more than
20 MeV lies between ~15% and ~30% (according to the MCNPX
simulations). The results are comparable to the discrepancies
observed in measurements at GSI (Silari et al., 2009), the CERF fa-
cility (Caresana et al., 2014) and the CERN Proton Synchrotron
(Manessi et al., 2014). In position 1a, however, a better agreement
was found between the LB 6411 measurement and the WENDI-2

Table 8
H*(10) results in mSv/(nA.h) from the BSS data unfolded with GRAVEL and MAXED
using the Bertini & Dresner guess spectrum, and e for position 1b e the BSS data
unfolded with GRAVEL and MAXED using the CEM03 guess spectrum.

Position Bertini guess spectrum CEM03 guess spectrum

GRAVEL MAXED GRAVEL MAXED

1b 1860 ± 223 1890 ± 227 1930 ± 232 1944 ± 233


1a 647 ± 78 648 ± 78 e e
1g 1.46 ± 0.18 1.30 ± 0.16 e e
2a 2.76 ± 0.33 2.78 ± 0.33 e e
Fig. 14. Event size spectra measured by the TEPC in positions 1a, 1b and 2b.
38 V. De Smet et al. / Radiation Measurements 99 (2017) 25e40

Table 9
Measurements with the WENDI-2, the LB 6411 and the TEPC, compared to the average H*(10) measured with the BSS. The results are expressed in mSv/(nA.h).

Position BSS H*(10) (Average) WENDI-2 LB 6411 TEPC

1b (1.9 ± 0.2)  103 (2.03 ± 0.09)  103 (1.23 ± 0.05)  103 (1.66 ± 0.07)  103
1a (6.5 ± 0.8)  102 (5.8 ± 0.2)  102 (5.5 ± 0.2)  102 (4.5 ± 0.2)  102
1g (1.4 ± 0.2)  100 (1.4 ± 0.1)  100 e e
2a (2.8 ± 0.3)  100 (2.7 ± 0.1)  100 (1.96 ± 0.09)  100 e
2c e (5.6 ± 0.2)  100 (4.1 ± 0.2)  100 e
2b e (3.9 ± 0.2)  100 (2.8 ± 0.1)  100 (3.4 ± 0.5)  100

and BSS data. According to the MCNPX simulations, neutrons with thermal neutron counter, which causes a very poor energy reso-
more than 20 MeV constitute only ~5% of the total neutron flux in lution in the thermal region since the response functions of the
this position. It seems that the LB 6411 can be used for area various spheres are nearly parallel at low energies (see Fig. 5).
monitoring of neutrons inside proton therapy facilities, but accu- Therefore, the shape of the guess spectrum has a strong influence
rate neutron H*(10) measurements behind the shielding and in on the thermal peak of the unfolded spectra.
“forward” positions inside the treatment room would actually
require the calculation of a position-specific calibration correction
factor. To determine such a correction factor, the shape of the local 4. Conclusions
neutron spectrum needs to be well characterszed and the response
function of the LB 6411 must be known up to 230 MeV. MCNPX shielding simulations of the FBTR were carried out with
The TEPC measures the dose equivalent H and was not cali- two different INC models (Bertini and CEM03) and were compared
brated for measuring H*(10) in these specific neutron fields. The to neutron measurements performed with an extended-range rem
results nevertheless agree with the other detector responses within meter WENDI-2, a conventional rem meter LB 6411, an extended-
at most two standard uncertainties in all positions (see Table 9). range BSS and a TEPC.
Compared to the WENDI-2 and BSS measurements, the TEPC results The measurements performed with the WENDI-2 agree within
tend to be lower by 20e30%. ~10% with the H*(10) rates measured with the BSS in the four
For positions 1a, 1b, 2a and 2b, the contribution of thermal considered positions, inside and outside the treatment room. It
neutrons to the total neutron H*(10) rate was calculated by folding confirms that, in these neutron fields, the WENDI-2 allows
the neutron spectra from the MCNPX simulations and the unfolded measuring H*(10) with satisfactory accuracy without applying any
BSS spectra with the fluence-to-H*(10) conversion coefficients field-specific calibration correction factor. The accuracy of the
(ICRP, 1996) between 0 and 2  107 MeV. The thermal H*(10) WENDI-2 is estimated to be within 10% for the other measurement
contributions calculated using the MCNPX spectra and the average positions as well (in the access maze and outside the FBTR), based
of the thermal H*(10) contributions derived from the unfolded BSS on a theoretical assessment using the WENDI-2 response function
spectra are compared to the 3He counter measurements in Table 10. and the simulated spectra.
It appears that, on average, the simulated thermal H*(10) contri- Measurements with an LB 6411, however, require a calibration
butions overestimate the 3He counter measurements by a factor ~2 correction factor in positions located either outside the shielding or
inside the treatment room, and by a factor of ~3e6 (depending on at forward angle with respect to the proton beam direction, to avoid
the selected physics models) outside the treatment room. The underestimating H*(10) by ~25e40%. As for the TEPC measure-
causes of these overestimates, especially inside the treatment ments, they are globally consistent with the other measurements,
room, are not well understood, but it is known that the thermal but the dose equivalents are lower by ~20e30% compared to the
neutron field in a given position may be strongly influenced by even H*(10) measured with the BSS and the WENDI-2.
relatively small objects that have not necessarily been modelled in In the MCNPX simulations, the choice of the physics models for
the simulations. Anyhow, this is not so important because the the proton and neutron interactions above 150 MeV has a signifi-
thermal contribution only represents a few percent of the total cant impact on the simulated neutron H*(10) outside the shielding.
neutron H*(10) rates. Moreover, in some areas inside the access For positions located at forward angles with respect to the proton
maze, where the neutron flux is largely thermal, the contribution of beam direction (e.g. positions 2ae2c), the Bertini and the CEM03
the thermal neutrons to H*(10) also remains relatively small models were identified as the most and the least conservative INC
compared to the contribution of neutrons of more than 0.1 MeV models, respectively. Inside the treatment room, the selected INC
(e.g. in position 1e, according to the simulations, no more than ~15% model essentially has a large impact on the high energy neutron
of H*(10) is due to thermal neutrons, which constitute ~70% of the flux, i.e. a limited fraction of the total flux, and thus the differences
neutron flux). As for the results derived from the unfolded BSS in H*(10) do not exceed factors of 1.2e1.3. The high-energy neu-
spectra, the thermal contribution to H*(10) is also larger than the trons in the source spectra however strongly determine the total
3
He counter measurements, by a factor of ~2 both inside and neutron flux outside the shielding, due to the propagation mech-
outside the treatment room. The BSS does not include a bare anisms of the INC inside the concrete. As a consequence, the dif-
ferences in H*(10) increase up to factors of 1.6e1.8 outside the

Table 10
Thermal contribution to the neutron H*(10) rate, in mSv/(nA.h), from the two MCNPX simulations (based on the Bertini & Dresner and CEM03 models, respectively), from the
unfolded BSS spectra (average result), and from the measurement with the 3He counter.
3
Position MCNPX BertiniThermal H*(10) MCNPX CEM03 Thermal H*(10) Average BSS resultsThermal H*(10) He counter

1a 25.99 ± 0.02 22.77 ± 0.02 25.4 ± 2.0 12 ± 3


1b 34.01 ± 0.02 27.64 ± 0.02 38.2 ± 3.0 17 ± 4
2a 0.2319 ± 0.0008 0.1430 ± 0.0006 0.0987 ± 0.0079 0.043 ± 0.011
2b 0.2806 ± 0.0009 0.1615 ± 0.0007 e 0.046 ± 0.012
V. De Smet et al. / Radiation Measurements 99 (2017) 25e40 39

shielding. Instrum. Methods Phys. Res. A 506, 250e303. http://dx.doi.org/10.1016/S0168-


9002(03)01368-8.
The comparison of the simulation results with the WENDI-2 and
Allison, J., Amako, K., Apostolakis, J., Araujo, H., Dubois, P.A., Asai, M., Barrand, G.,
BSS measurements for positions inside the treatment room showed Capra, R., Chauvie, S., Chytracek, R., Cirrone, G.A.P., Cooperman, G., Cosmo, G.,
that the Bertini & Dresner models reproduce the global neutron Cuttone, G., Daquino, G.G., Donszelmann, M., Dressel, M., Folger, G., Foppiano, F.,
production in the water phantom relatively well, whereas the Generowicz, J., Grichine, V., Guatelli, S., Gumplinger, P., Heikkinen, A.,
Hrivnacova, I., Howard, A., Incerti, S., Ivanchenko, V., Johnson, T., Jones, F., Koi, T.,
CEM03 model underestimates it by a factor of ~1.3. At the most- Kokoulin, R., Kossov, M., Kurashige, H., Lara, V., Larsson, S., Lei, F., Link, O.,
forward angle, however, the Bertini model (unlike the CEM03 Longo, F., Maire, M., Mantero, A., Mascialino, B., McLaren, I., Lorenzo, P.M.,
model) seems to overestimate the production of neutrons with Minamimoto, K., Murakami, K., Nieminen, P., Pandola, L., Parlati, S., Peralta, L.,
Perl, J., Pfeiffer, A., Pia, M.G., Ribon, A., Rodrigues, P., Russo, G., Sadilov, S.,
energies above 100 MeV. In the Bertini & Dresner simulation, this Santin, G., Sasaki, T., Smith, D., Starkov, N., Tanaka, S., Tcherniaev, E., Tome, B.,
aspect might contribute in a significant way to the overestimate on Trindade, A., Truscott, P., Urban, L., Verderi, M., Walkden, A., Wellisch, J.P.,
the neutron H*(10) outside the shielding at forward angle. Williams, D.C., Wright, D., Yoshida, H., 2006. GEANT4 developments and ap-
plications. IEEE Trans. Nucl. Sci. 53, 270e278.
Compared to the WENDI-2 measurements, overestimates by Aza, E., Dinar, N., Manessi, G.P., Silari, M., 2015. A Bonner Sphere Spectrometer for
factors of 2e3 were observed on the neutron H*(10) outside the pulsed fields. Radiat. Prot. Dosim. http://dx.doi.org/10.1093/rpd/ncv180.
shielding when using the Bertini & Dresner models. These results Bertini, H.W., 1969. Intranuclear-cascade calculation of the secondary nucleon
spectra from nucleon-nucleus interactions in the energy range 340 to 2900
are comparable to those of a similar study conducted at Essen (De MeV and comparisons with experiment. Phys. Rev. C 188, 1711.
Smet et al., 2014b). With the CEM03 model, overestimates of up Bertini, H.W., 1963. Low-energy intranuclear cascade calculation. Phys. Rev. C 131,
to a factor of 1.7 were obtained outside the shielding, in spite of the 1801.
Birattari, C., Dimovasili, E., Mitaroff, A., Silari, M., 2010. A Bonner Sphere Spec-
underestimates observed inside the treatment room. Both simu-
trometer with extended response matrix. Nucl. Instrum. Methods Phys. Res. A
lations are thus conservative with respect to the neutron fluxes 620, 260e269. http://dx.doi.org/10.1016/j.nima.2010.04.033.
transmitted through the concrete walls. Although uncertainties on Birattari, C., Esposito, A., Ferrari, A., Pelliccioni, M., Silari, M., 1992. A neutron survey
the concrete composition and density might contribute to this meter with sensitivity extended up to 400 MeV. Radiat. Prot. Dosim. 44,
193e197.
conservativeness, they are unlikely to be the single cause of it (De Birattari, C., Ferrari, A., Nuccetelli, C., Pelliccioni, M., Silari, M., 1990. An extended
Smet, 2016). Future work should be dedicated to investigating the range neutron rem counter. Nucl. Instrum. Methods Phys. Res. A 297, 250e257.
influence of other sources of uncertainties in the simulations, such http://dx.doi.org/10.1016/0168-9002(90)91373-J.
Bo€hlen, T.T., Cerutti, F., Chin, M.P.W., Fasso , A., Ferrari, A., Ortega, P.G., Mairani, A.,
as e.g. uncertainties on the cross-sections for neutron (and proton) Sala, P.R., Smirnov, G., Vlachoudis, V., 2014. The FLUKA code: developments and
interactions inside the concrete and uncertainties on the high- challenges for high energy and medical applications. Nucl. Data Sheets 120,
energy end of the double-differential neutron production cross- 211e214. http://dx.doi.org/10.1016/j.nds.2014.07.049.
Booth, T.E., 1985. A Sample Problem for Variance Reduction in MCNP. LA-10363-MS.
section of protons in water. Briesmeister, J.F., 1997. MCNP d a General Monte Carlo N-particle Transport Code,
Version 4B. LA-12625-M. Los Alamos.
Funding Burgkhardt, B., Fieg, G., Klett, A., Plewnia, A., Siebert, B.R.L., 1997. The neutron flu-
ence and H*(10) response of the new LB 6411 remcounter. Radiat. Prot. Dosim.
70, 361e364.
This research project has been partly supported by: the FIRST Caresana, M., Silari, M., Helmecke, M., Kubancak, J., Manessi, G.P., Ott, K.,
project ‘FREDONE’ funded by the Re gion Wallonne (Belgium) Scherpelz, R., 2014. Instrument intercomparison in the high-energy mixed field
at the CERN-EU reference field (CERF) facility. Radiat. Prot. Dosim. 161, 67e72.
(1117285). a Marie Curie Initial Training Network Fellowship of the Chadwick, M.B., Oblo zinský, P., Herman, M., Greene, N.M., McKnight, R.D.,
European Community's Seventh Framework Programme under Smith, D.L., Young, P.G., MacFarlane, R.E., Hale, G.M., Frankle, S.C., Kahler, A.C.,
contract number PITN-GA-2011-289198-ARDENT. Kawano, T., Little, R.C., Madland, D.G., Moller, P., Mosteller, R.D., Page, P.R.,
Talou, P., Trellue, H., White, M.C., Wilson, W.B., Arcilla, R., Dunford, C.L.,
Mughabghab, S.F., Pritychenko, B., Rochman, D., Sonzogni, A.A., Lubitz, C.R.,
Acknowledgements Trumbull, T.H., Weinman, J.P., Brown, D.A., Cullen, D.E., Heinrichs, D.P.,
McNabb, D.P., Derrien, H., Dunn, M.E., Larson, N.M., Leal, L.C., Carlson, A.D.,
Block, R.C., Briggs, J.B., Cheng, E.T., Huria, H.C., Zerkle, M.L., Kozier, K.S.,
The authors would like to thank Eric Malotaux (IBA) and the Courcelle, A., Pronyaev, V., van der Marck, S.C., 2006. ENDF/B-VII.0: next gen-
Westdeutsches Protonentherapiezentrum Essen for having granted eration evaluated nuclear data library for nuclear science and technology. Nucl.
them the opportunity to perform this experiment. Data Sheets 107, 2931e3060. http://dx.doi.org/10.1016/j.nds.2006.11.001.
Conte, V., Moro, D., Grosswendt, B., Colautti, P., 2013. Lineal energy calibration of
mini tissue-equivalent gas-proportional counters (TEPC). AIP Conf. Proc. 1530,
References 171e178. http://dx.doi.org/10.1063/1.4812920.
De Smet, V., 2016. Neutron Measurements in a Proton Therapy Facility and Com-
Agosteo, S., Bedogni, R., Caresana, M., Charitonidis, N., Chiti, M., Esposito, A., parison with Monte Carlo Shielding Simulations. PhD thesis. Universite  Libre de
Ferrarini, M., Severino, C., Silari, M., 2012. Characterization of extended range Bruxelles.
Bonner Sphere Spectrometers in the CERF high-energy broad neutron field at De Smet, V., Gerardy, I., Stichelbaut, F., Tolo, S., 2014a. MCNPX simulations of the
CERN. Nucl. Instrum. Methods Phys. Res. A 694, 55e68. http://dx.doi.org/ response of the extended-range rem meter WENDI-2. Suppl Nucl. Technol.
10.1016/j.nima.2012.06.055. Radiat. Prot. 29, 25e30. http://dx.doi.org/10.2298/NTRP140SS25S.
Agostinelli, S., Allison, J., Amako, K., Apostolakis, J., Araujo, H., Arce, P., Asai, M., De Smet, V., Stichelbaut, F., Vanaudenhove, T., Mathot, G., De Lentdecker, G.,
Axen, D., Banerjee, S., Barrand, G., Behner, F., Bellagamba, L., Boudreau, J., Dubus, A., Pauly, N., Gerardy, I., 2014b. Neutron H*(10) inside a proton therapy
Broglia, L., Brunengo, A., Burkhardt, H., Chauvie, S., Chuma, J., Chytracek, R., facility: comparison between Monte Carlo simulations and WENDI-2 mea-
Cooperman, G., Cosmo, G., Degtyarenko, P., Dell’Acqua, A., Depaola, G., surements. Radiat. Prot. Dosim. 161, 417e421.
Dietrich, D., Enami, R., Feliciello, A., Ferguson, C., Fesefeldt, H., Folger, G., Dresner, L., 1961. EVAP - a Fortran Program for Calculating the Evaporation of
Foppiano, F., Forti, A., Garelli, S., Giani, S., Giannitrapani, R., Gibin, D., Gomez Various Particle Evaporation from Excited Nuclei. ORNL-TM-196.
Cadenas, J.J., Gonzalez, I., Gracia Abril, G., Greeniaus, G., Greiner, W., Grichine, V., , A., Ranft, J., 2005. Fluka: a Multi-particle Transport Code.
Ferrari, A., Sala, P.R., Fasso
Grossheim, A., Guatelli, S., Gumplinger, P., Hamatsu, R., Hashimoto, K., Hasui, H., CERN 2005-010, INFN TC_05/11, SLAC-R-773.
Heikkinen, A., Howard, A., Ivanchenko, V., Johnson, A., Jones, F.W., Kallenbach, J., Folger, G., Ivanchenko, V.N., Wellisch, J.P., 2004. The binary cascade. Eur. Phys. J. A
Kanaya, N., Kawabata, M., Kawabata, Y., Kawaguti, M., Kelner, S., Kent, P., 21, 407e417. http://dx.doi.org/10.1140/epja/i2003-10219-7.
Kimura, A., Kodama, T., Kokoulin, R., Kossov, M., Kurashige, H., Lamanna, E., Harano, H., Nolte, R., 2011. Quasi-monoenergetic high-energy neutron standards
Lampen, T., Lara, V., Lefebure, V., Lei, F., Liendl, M., Lockman, W., Longo, F., above 20 MeV. Metrologia 48, S292eS303. http://dx.doi.org/10.1088/0026-
Magni, S., Maire, M., Medernach, E., Minamimoto, K., Mora de Freitas, P., 1394/48/6/S06.
Morita, Y., Murakami, K., Nagamatu, M., Nartallo, R., Nieminen, P., Nishimura, T., ICRP, 2010. Conversion Coefficients for Radiological Protection Quantities for
Ohtsubo, K., Okamura, M., O'Neale, S., Oohata, Y., Paech, K., Perl, J., Pfeiffer, A., External Radiation Exposures, 116. ICRP Publication.
Pia, M.G., Ranjard, F., Rybin, A., Sadilov, S., di Salvo, E., Santin, G., Sasaki, T., ICRP, 1996. Conversion Coefficients for Use in Radiological Protection against
Savvas, N., Sawada, Y., Scherer, S., Sei, S., Sirotenko, V., Smith, D., Starkov, N., External Radiation, 74. ICRP Publication.
Stoecker, H., Sulkimo, J., Takahata, M., Tanaka, S., Tcherniaev, E., Safai Tehrani, E., ICRU, 1998. Clinical Proton Dosimetry - Part I: beam Production, Beam Delivery and
Tropeano, M., Truscott, P., Uno, H., Urban, L., Urban, P., Verderi, M., Walkden, A., Measurement of Absorbed Dose, ICRU Report, 59.
Wander, W., Weber, H., Wellisch, J.P., Wenaus, T., Williams, D.C., Wright, D., ICRU, 1980. Quantities and Units for Use in Radiation Protection, ICRU Report, 33.
Yamada, T., Yoshida, H., Zschiesche, D., 2003. GEANT4-a simulation toolkit. Nucl. €gerhofer, L., Feldbaumer, E., Forkel-Wirth, D., Theis, C., Vincke, H., Iwamoto, Y.,
Ja
40 V. De Smet et al. / Radiation Measurements 99 (2017) 25e40

Hagiwara, M., Satoh, D., Iwase, H., Yashima, H., Matsumoto, T., Masuda, A., field, part I: Monte Carlo simulations. Radiat. Meas. 44, 649e659. http://
Nishiyama, J., Nakamura, T., Sato, T., Nakane, Y., Nakashima, H., Sakamoto, Y., dx.doi.org/10.1016/j.radmeas.2009.03.029.
Tamii, A., Hatanaka, K., 2011. Characterization of the WENDI-II rem counter for Rühm, W., Mares, V., Pioch, C., Agosteo, S., Endo, A., Ferrarini, M., Rakhno, I.,
its application at MedAustron. Prog. Nucl. Sci. Technol. 2, 258e262. Rollet, S., Satoh, D., Vincke, H., 2014. Comparison of Bonner Sphere responses
€gerhofer, L., Feldbaumer, E., Theis, C., Roesler, S., Vincke, H., 2012. A new method
Ja calculated by different Monte Carlo codes at energies between 1 MeV and 1 GeV
to calculate the response of the WENDI-II rem counter using the FLUKA Monte e potential impact on neutron dosimetry at energies higher than 20 MeV.
Carlo code. Nucl. Instrum. Methods Phys. Res. A 691, 81e85. Radiat. Meas. 67, 24e34. http://dx.doi.org/10.1016/j.radmeas.2014.05.006.
Manessi, G.P., Aza, E., Caresana, M., Cassell, C., Colombo, V., Damjanovic, S., Sannikov, A.V., Savitskaya, E.N., 1997. Ambient dose equivalent conversion factors
Gilardoni, S., Pangallo, M., Perrin, D., 2014. Comparison of the performance of for high energy neutrons based on the ICRP 60 recommendations. Radiat. Prot.
different instruments in the stray neutron field around the CERN proton syn- Dosim. 70, 383e386.
chrotrons. Radiat. Prot. Dosim. 161, 190e195. Sato, T., Niita, K., Matsuda, N., Hashimoto, S., Iwamoto, Y., Noda, S., Ogawa, T.,
Mashnik, S.G., Gudima, K.K., Prael, R.E., Sierk, A.J., Baznat, M.I., Mokhov, N.V., 2008. Iwase, H., Nakashima, H., Fukahori, T., Okumura, K., Kai, T., Chiba, S., Furuta, T.,
CEM03.03 and LAQGSM03.03 Event Generators for the MCNP6, MCNPX, and Sihver, L., 2013. Particle and heavy ion transport code system. J. Nucl. Sci.
MARS15 Transport Codes. http://dx.doi.org/10.1016/j.nimb.2010.09.005. LA-UR- Technol. 50, 913e923. http://dx.doi.org/10.1080/00223131.2013.814553.
08e2931. Satoh, D., Maeda, Y., Tameshige, Y., Nakashima, H., Shibata, T., Endo, A., Tsuda, S.,
Mashnik, S.G., Sierk, A.J., Bersillon, O., Gabriel, T., 1998. Cascade-exciton model Sasaki, M., Maekawa, M., Shimizu, Y., Yamazaki, M., Katayose, T., Niita, K., 2012.
analysis of proton induced reactions from 10 MeV to 5 GeV. Nucl. Instrum. Shielding study at the Fukui prefectural hospital proton therapy center. J. Nucl.
Methods Phys. Res. A 414, 68e72. Sci. Technol. 49, 1097e1109. http://dx.doi.org/10.1080/00223131.2012.730900.
Matzke, M., 2003. Unfolding procedures. Radiat. Prot. Dosim. 107, 155e174. Silari, M., Agosteo, S., Beck, P., Bedogni, R., Cale, E., Caresana, M., Domingo, C.,
Matzke, M., 1994. Unfolding of Pulse Height Spectra: the HEPRO Program System. Donadille, L., Dubourg, N., Esposito, A., Fehrenbacher, G., Ferna ndez, F.,
Physikalisch-Technische Bundesanstalt, Braunschweig. PTB-N-19. Ferrarini, M., Fiechtner, A., Fuchs, A., García-Fuste , M.J., Golnik, N., Gutermuth, F.,
Mayer, S., Forkel-Wirth, D., Fuerstner, M., Menzel, H.G., Mueller, M.J., Perrin, D., Khurana, S., Klages, T., Latocha, M., Mares, V., Mayer, S., Radon, T., Reithmeier, H.,
Theis, C., Vincke, H., 2007. Response of neutron detectors to high-energy mixed Rollet, S., Roos, H., Rühm, W., Sandri, S., Schardt, D., Simmer, G., Spurný, F.,
radiation fields. Radiat. Prot. Dosim. 125, 289e292. http://dx.doi.org/10.1093/ Trompier, F., Villa-Grasa, C., Weitzenegger, E., Wiegel, B., Wielunski, M.,
rpd/ncm182. Wissmann, F., Zechner, A., Zielczyn  ski, M., 2009. Intercomparison of radiation
McConn Jr., R.J., Gesh, C.J., Pagh, R.T., Rucker, R.A., Williams III, R.G., 2011. Com- protection devices in a high-energy stray neutron field. part III: instrument
pendium of Material Composition Data for Radiation Transport Modeling - response. Radiat. Meas. 44, 673e691. http://dx.doi.org/10.1016/
Revision 1. PIET-43741eTM963, PNNL-15870 Rev. 1. j.radmeas.2009.05.005.
Olsher, R.H., Hsu, H.H., Beverding, A., Kleck, J.H., Casson, W.H., Vasilik, D.G., Thomas, D.J., 2014. The system of radiation protection for neutrons: does it fit the
Devine, R.T., 2000. WENDI: an improved neutron rem meter. Health Phys. 79, purpose? Radiat. Prot. Dosim. 161, 3e10.
170e181. Trellue, H.R., Little, R.C., Lee, M.B., 2008. New ACE-formatted Neutron and Proton
Olsher, R.H., McLean, T.D., 2008. High-energy response of the PRESCILA and Libraries Based on ENDF/B-VII.0. LA-UR-08e1999.
WENDI-II neutron rem meters. Radiat. Prot. Dosim. 130, 510e513. http:// Trellue, H.R., Little, R.C., White, M.C., MacFarlane, R.E., Kahler, A.C., 2009. ENDF70: a
dx.doi.org/10.1093/rpd/ncn092. continuous-energy MCNP neutron data library based on ENDF/B-VII.0. Nucl.
Pelowitz, D.B., 2011. MCNPX User’s Manual Version 2.7.0. LA-CP-11e00438. Technol. 168, 832e836.
Perez-Andújar, A., Deluca, P.M., Thornton, A.F., Fitzek, M., Hecksel, D., Farr, J., 2012. Vanaudenhove, T., 2014. Shielding Study against High-energy Neutrons Produced in
Microdosimetric measurements for neutron-absorbed dose determination a Proton Therapy Facility by Means of Monte Carlo Codes and On-site Mea-
during proton therapy. Radiat. Prot. Dosim. 151, 365e373. http://dx.doi.org/ surements. PhD thesis. Universite  Libre de Bruxelles.
10.1093/rpd/ncs002. Vanaudenhove, T., Stichelbaut, F., Dubus, A., Pauly, N., De Smet, V., 2014. Monte
Pioch, C., Mares, V., Rühm, W., 2010. Influence of Bonner Sphere response functions Carlo calculations with MCNPX and GEANT4 for general shielding study e
above 20 MeV on unfolded neutron spectra and doses. Radiat. Meas. 45, application to a proton therapy center. Prog. Nucl. Sci. Technol. 4, 422e426.
1263e1267. http://dx.doi.org/10.1016/j.radmeas.2010.05.007. Waker, A.J., 1995. Principles of experimental microdosimetry. Radiat. Prot. Dosim.
PTW, 2008. User Manual Bragg Peak Chambers Type 34070, 34073 and 34080. 61, 297e308.
D797.131.00/04. Waker, A.J., Aslam, 2010. A preliminary study of the performance of a novel design
Reginatto, M., 2009. What can we learn about the spectrum of high-energy stray of multi-element tissue equivalent proportional counter for neutron moni-
neutron fields from Bonner Sphere measurements? Radiat. Meas. 44, 692e699. toring. Radiat. Meas. 45, 1309e1312. http://dx.doi.org/10.1016/
http://dx.doi.org/10.1016/j.radmeas.2009.04.005. j.radmeas.2010.05.025.
Reginatto, M., 2004. Manual GRAVEL MAXED - UMG33 Package. Weisskopf, V., 1937. Statistics and nuclear reactions. Phys. Rev. 52, 295e302.
Reginatto, M., Goldhagen, P., 1999. MAXED, a computer code for maximum entropy Wiegel, B., Agosteo, S., Bedogni, R., Caresana, M., Esposito, A., Fehrenbacher, G.,
deconvolution of multisphere neutron spectrometer data. Health Phys. 77, Ferrarini, M., Hohmann, E., Hranitzky, C., Kasper, A., Khurana, S., Mares, V.,
579e583. http://dx.doi.org/10.1097/00004032-199911000-00012. Reginatto, M., Rollet, S., Rühm, W., Schardt, D., Silari, M., Simmer, G.,
Reginatto, M., Goldhagen, P., Neumann, S., 2002. Spectrum unfolding, sensitivity Weitzenegger, E., 2009. Intercomparison of radiation protection devices in a
analysis and propagation of uncertainties with the maximum entropy decon- high-energy stray neutron field, part II: Bonner Sphere Spectrometry. Radiat.
volution code MAXED. Nucl. Instrum. Methods Phys. Res. A 476, 242e246. Meas. 44, 660e672. http://dx.doi.org/10.1016/j.radmeas.2009.03.026.
http://dx.doi.org/10.1016/S0168-9002(01)01439-5. X-5 Monte Carlo Team, 2003. MCNP - a General Monte Carlo N-particle Transport
Rollet, S., Agosteo, S., Fehrenbacher, G., Hranitzky, C., Radon, T., Wind, M., 2009. Code, Version 5-Volume I: Overview and Theory. LA-UR-03e1987.
Intercomparison of radiation protection devices in a high-energy stray neutron

You might also like