Measurement of detection efficiency using
Pb-scintillating fiber sampling KLOE calorimeter
for neutrons between 22 and 174 MeV
M.Anellia, G.Battistonib, S.Bertoluccia, C.Binic, P.Branchinid, C.Curceanua, G.De Zorzic, A.Di Domenicoc,
B.Di Miccod, A.Ferrarie, S.Fiore, P.Gauzzic, S.Giovannellaa, F.Happachera, M.Iliescua,f, M.Martinia,h,
S.Miscettia, F.Nguyend, A.Passerid, A.Prokofievg, P.Salab, B.Sciasciaa,1, F.Sirghia
(a) Laboratori Nazionali di Frascati, INFN, Italy,
(b) Sezione INFN di Milano, Italy,
(c) Universit` a degli Studi “La Sapienza” e Sezione INFN di Roma, Italy,
(d) Universit` a degli Studi “Roma Tre” e Sezione INFN di Roma Tre, Italy,
(e) Fondazione CNAO, Milano, Italy,
(f) IFIN-HH, Bucharest, Romania,
(g) The Svedberg Laboratory, Uppsala University, Sweden.
(h) Dipartimento di Energetica dell’Universit´ a “La Sapienza”,Roma, Italy.
(1) Corresponding author: email@example.com
Abstract—We exposed a prototype of the high-sampling lead-
scintillating fiber KLOE calorimeter to neutron beam of 21,
46, and 174 MeV, provided by the The Svedberg Laboratory
(TSL), Uppsala, to study its neutron detection efficiency. This
has been found larger than what what expected considering the
scintillator thickness of the KLOE prototype only. We checked the
reliability of our method, measuring also the neutron detection
efficiency of a 5 cm thick NE110 scintillator. Our results prove
the existence of a contribution from passive material to neutron
detection efficiency, in a high-sampling calorimeter configuration.
The origin of the efficiency enhancement has been studied using
the FLUKA Monte Carlo code. We present the TSL beam test
results and the reasons for such enhancement, which allows to
develop compact, inexpensive, fast, and highly efficient neutron
Index Terms—calorimeter, neutron detection efficiency, scintil-
Detection of neutrons with energies from a few to hundreds
MeV is usually performed with organic scintillators. The high
concentration of hydrogen atoms provides a proton target and
the elastic scattering of neutrons produces recoil protons which
produce a visible response. Typical efficiency is of the order of
1 % per cm of scintillator thickness . On the other hand the
extended range rem counters used in radiation protection ,
 are based on the idea of adding one layer of a medium-
high Z material to the organic scintillator. This enhances the
neutron response through inelastic processes that result in an
abundant production of secondary neutrons.
The KLOE calorimeter  is a high sampling lead scintillat-
ing fiber calorimeter. Energy and time resolutions are σE/E =
5.7%/?E(GeV) and σt = 57 ps/?E (GeV) ⊕ 100 ps.
The calorimeter was primarily designed to detect photons, the
detection efficiency is ∼ 90% at E = 20 MeV and reaches
100% above 70 MeV.
A measurement  performed at KLOE using charged kaon
interactions in the apparatus walls showed high efficiency
for neutron of about 20 MeV of the KLOE calorimeter. The
average efficiency measured was between three and four times
higher than the one expected if the response were only due to
the equivalent amount of scintillator in the calorimeter. This
result was confirmed also by the official Monte Carlo simula-
tion of the KLOE experiment. To understand the underlying
physical mechanisms which produce this difference, and to
make a more systematic study of the calorimeter response, we
planned a test beam to expose the calorimeter to a dedicated
neutron beam and we developed a complete simulation of the
detector geometry and response.
Apart from the possibility to develop compact, inexpensive,
fast and highly efficient neutron counters, this work has also
been motivated by the prospects of the search for deeply bound
kaonic nuclei  and of the neutron electromagnetic form
factor measurements in the time-like region . These are
two fundamental items for the DAΦNE upgrade under study
at the Laboratori Nazionali di Frascati of INFN (LNF). In
both cases the neutron detection efficiency is a key point for
II. MEASUREMENT OF THE NEUTRON DETECTION
EFFICIENCY AT TSL
The KLOE calorimeter prototype used in this measurement
is composed of ∼ 200 layers of 1 mm diameter blue scin-
tillating fibers glued inside grooved lead layers of 0.5 mm
thickness. The final structure has a fiber:lead:glue volume
ratio of 48:42:10 resulting in a density of ∼ 5 g/cm3. The
total external dimensions are 13 × 24 × 65 cm3where the
second value is the calorimeter depth and the third one is
the fiber length. The calorimeter is readout at both fiber ends
in order to reconstruct this coordinate by time difference.
Fig. 1. Experimental setup used during test beam. (1) Calorimeter prototype,
(2) beam position monitor and (3) NE110 reference counter.
The readout is organized in four planes along the calorimeter
depth for the first 16.8 cm. Each plane is then divided in 3
columns along the horizontal coordinate originating cells of
4.2×4.2 cm2. Larger readout elements are used in the rear
part of the calorimeter. The small elements are coupled to
1-1/8 inch diameter standard bialkali photomultipliers, PM,
through light guides.
When running with neutron beam, the calorimeter was
positioned with fibers running vertically, the lower (upper)
end is called side A (B). Each PM signal has been split in
three replicas. The first two signals allow measurements of
charge and time while the third one is used to summing up
the PM charges of the first four planes of each side. The
discriminated signals obtained, SA and SB, are combined in
overlap coincidence, SA · SB, to trigger on the calorimeter.
A reference counter for the efficiency measurement was
built with a 5 cm thick bulk of NE110 organic scintillator,
of transversal dimensions 10×20 cm2, by coupling it at the
two ends to two EMI9814 PM’s.
A “beam position monitor” is also used. This detector was
made by an array of seven scintillating counters 1 cm thick to
check the beam shape during run.
The experimental setup is shown in Fig. 1, where we have
the calorimeter prototype, the reference counter and the beam
position monitor used to detect beam dimensions.
When running with the beam, the scintillator was positioned
with its longest dimension along the horizontal coordinate. To
trigger on the scintillator, the PM signals, S1 and S2, were
discriminated and an overlap coincidence, S1·S2, was formed.
We ran our experiment at the “The Svedberg Laboratory”
(TSL) high-energy neutron facility . At this facility, protons
from the Gustaf Werner cyclotron were directed on a7Li target
generating a neutron beam which is then geometrically shaped
by an iron collimator block with a 2 cm diameter cylindrical
hole. The energy and angular distributions of neutrons are de-
fined by the differential cross section of the7Li(p,n) reaction at
forward angles . The neutron energy spectrum is dominated
by a peak at a few MeV below the primary proton energy,
Fig. 2.Neutrons energy spectrum at the TSL facility.
and a long tail down to thermal neutrons (see Fig. 2). In our
settings, we performed two different test beams, using high
energy neutrons (178.7 MeV) and low energy neutrons (21
and 46 MeV). The neutron beam time structure was in phase
with the cyclotron RF which had a period, TRF, of ∼ 45 ns.
The beam was emerging from the collimator at 3 m distance
from the target. In order to ensure full beam acceptance we
ran at a distance of 5.1 m from the target and centered the
beam in the middle of our detectors. Very low intensity neutron
beams, from 1.5 kHz/cm2to 6 kHz/cm2, have been required
in order to minimize the probability of counting more than a
neutron per event. The neutron rate has been measured with a
ionization-chamber monitor (ICM), with an absolute accuracy
estimated at the level of 10 %.
All discrimination and trigger signals were formed with
NIM logic while the DAQ was based on VME standard. Data
have been acquired with a simplified version of the KLOE
data acquisition system, composed by 40+40 ADC and TDC
channels and able to write on disk at a rate of 1.7 kHz. The
effective ADC gate was 300 ns wide, while the TDC worked in
common start mode with the start provided by the trigger and
the stop given by the discriminated signals delayed of 150 ns
by a monostable inside the board. The trigger signal was either
SA·SB or S1·S2 after phase locking with the cyclotron RF
replica. Typical runs consist of ∼ 0.5 − 1.5 ×106events.
The detector efficiency to the overall neutron spectrum, ǫD,
has been determined taking the ratio of the counted detector
rate, RD, with the rate of the neutron beam estimated by the
ICM, Rn. For a given trigger threshold, assuming full beam
acceptance and no background, RD is the ratio between the
acquired rate, RDAQ, and the fraction of DAQ live time, Flive.
The incoming neutron rate, Rn, is instead given by:
Rn(kHz) = 9.09 π · RICM· Fpeak
where RICM is the ICM rate in Hz counted in each run and
Fpeakis the fraction of neutrons at peak energy.
calorimeter cell for the first calorimeter plane with 174 MeV neutrons.
Dots are data, histograms are the simulation (solid) and background (green)
contributions. (left) central cell, (right) outer cell.
Data-Monte Carlo comparison of time of flight in the central
III. CALIBRATION AND MEASURED EFFICIENCIES
For the scintillator, the horizontal scale has been calibrated
fitting the energy spectra obtained with a reference β source.
For the calorimeter, each cell has been calibrated using min-
imum ionizing particles and then applying the energy scale
determined with electron and photon beams  in the same
The effective trigger thresholds cut-off have been evaluated
from the data looking at the charge integrated by all the cells
of the calorimeter at different thresholds. The ratio of these
distributions normalized to the tail integrals has been fit with a
Fermi-Dirac functions to extract the effective threshold value.
A clustering procedure is used to reconstruct the timing in
the whole calorimeter and derive the kinetic energy of the
primary neutron from time of flight, TOF. A first data-Monte
Carlo comparison of TOF has been made at cell level. In
the cells of the central column more than 90 % of the beam
is contained. Data reconstructed clusters, with a single fired
cell, show a ratio lateral/central fired cells higher then what
expected from the simulation. Moreover, lateral cells show
a flatter time distribution compared with Monte Carlo. We
interpreted this effect as a presence of background due to low
energy neutrons that form an halo around the central beam
core. An overall good agreement is observed when taking
into account also the shape of the halo contribution; this is
obtained from data looking at the TOF distribution of the outer
cells of the calorimeter where the background contamination
is evident. We have estimated, for the run at 174 MeV, that
halo events contribute to ∼ 30% of the total number of events.
In Fig. 3, we show the TOF distribution for the central and
the outer cells of the first calorimeter plane.
Concerning low energy run, we are still completing the TOF
study to evaluate the correct halo fraction. As a preliminary
evaluation, we rely on halo measurement carried out by TSL
beam experts. They performed a scan of the area near the
collimator using a fission monitor counter. This measurement
has been confirmed by our background counters. By integrat-
ing over the whole calorimeter plane, we estimate a fraction
of halo contamination: Fh= (20 ± 10)%.
The measured efficiencies for the scintillator and for the
calorimeter are shown in Fig. 4, and Figs.5 and 6, as a function
of the trigger threshold in MeV, after correcting for the halo
En = 174 MeV - R = 3.0 kHz/cm2
En = 174 MeV - R = 1.5 kHz/cm2
En = 46.5 MeV
En = 21.8 MeV
Fig. 4. Dependence of ǫscinton the applied trigger threshold. The horizontal
scale is in MeV of electron equivalent energy. The systematics is 10 % on
both horizontal and vertical scale. A comparison with available measurements
in literature  is also shown.
The horizontal scale is in MeV set for electron response.
Dependence of ǫcaloon the trigger threshold (run at En174 MeV).
fraction. In Figs.5 and 6, the calorimeter efficiency is also
compared with the expected result obtained considering the
scintillator equivalent thickness.
These results show that at the lowest trigger threshold
the neutron detection efficiency of the calorimeter ranges
from 28 % to 33 %, depending on the beam intensity. For
comparison, the efficiency of the 5 cm thick NE110 scintillator
ranges from 4 % to 10 %, for values of the trigger threshold
below 5 MeV of electron equivalent energy, in good agreement
with the available measurements in literature. This indicates
and En∼46 MeV). The horizontal scale is in MeV set for electron response.
Dependence of ǫcaloon the trigger threshold (run at En∼21 MeV
that the measured calorimeter efficiency is sizeably enhanced
with respect to the expected 8−10% based on the amount of
IV. MONTE CARLO SIMULATION AND COMPARISON WITH
The Monte Carlo code FLUKA ,  has been used
for a detailed simulation of the calorimeter structure. The
TSL experimental beam-line, from the neutron source to the
collimated beam, has been also simulated, in order to have a
reliable characterization of the neutron beam impinging on the
detector (see Fig. 7). Using the tool “LATTICE” the whole
calorimeter module has been designed. All the compounds
have been carefully simulated: for the fibers, an average
density between cladding and core has been used, for the glue
we have taken into account the right fraction of epoxy resin
and hardener. FLUKA computes the energy deposits in the
scintillating fibers, taking into account the signal saturation
due to the Birks law. For each energy deposit, the average
number of photoelectrons is estimated and then attenuated to
the calorimeter ends with the proper attenuation length. The
photoelectron statistics and the generation of the discriminated
signal are also simulated, while the trigger effect has not yet
The primary reason for the observed efficiency enhancement
appears to be the huge inelastic production of neutrons on
the lead planes. For neutrons in the high energy peak (175
MeV), the probability to have an inelastic interaction is 31.4 %
on the lead, compared to 7.0 % on the fiber and 2.2 % on
the glue. The secondary particles generated in such inelastic
interactions are on average 5.4 per event, counting only the
secondary neutrons above 19.6 MeV. Among the produced
secondaries, 62 % are neutrons, 27 % photons, 7 % protons
while the remaining 4 % are nuclear fragments. Typical
inelastic reactions on lead are:
From the top: at the source, at the collimator exit and on the calorimeter
Neutron energy spectra as computed with the FLUKA simulation.
- n Pb → xn + yγ + Pb,
- n Pb → xn + yγ + p + residual nucleus,
- n Pb → xn + yγ + 2p + residual nucleus.
Low-energy neutrons (below 19.6 MeV) are transported in
FLUKA with a multi-group algorithm, that uses a neutron
cross-section library derived from the most recently evaluated
data. These secondary neutrons give also a sizeable contribu-
tion to the response: due to the larger inelastic cross section the
neutron shower-like effect increases and originates on average
about 100 secondaries per event, out of which ∼ 5 protons and
∼ 1 photon directly contribute in generating a visible response.
The high sampling frequency of the calorimeter appears to
be a crucial point in the efficiency enhancement. First of all,
the protons and the electromagnetic energy produced on lead
in the inelastic processes can be detected by the nearby fibers
down to very low energies. Moreover, secondary neutrons are
produced on following lead planes with decreasing energies,
thus having larger probabilities to produce ionizing particles on
the nearby fibers. The isotropic distributions, which character-
ize the inelastic processes, also play a role: the back scattered
neutrons contribute to increase the collision density in the first
calorimeter planes, so containing the neutron shower depth.
A good agreement between data and simulation is observed
when taking into account also the halo contribution as shown
in Fig. 3 for the TOF and in Fig. 8 for the collected charge.
In Fig. 9 we show the dependence of ǫcalo as a function
of the trigger threshold compared to the FLUKA simulation
after correcting RDAQfor the halo contribution estimated with
the TOF. The comparison is done for the run at 174 MeV. A
pretty good data/simulation agreement is observed. Compari-
son between data and Monte Carlo for the low energy run is
0 50100 150200250
Halo contribution has been included in the simulation (bottom plot).
Comparison between data and Monte Carlo for the collected charge.
as function of the applied trigger threshold for the run at 174 MeV.
ǫcalocomparison between data (bullet) and simulation (open circle)
V. CONCLUSION AND PROSPECTS
Using an high sampling Pb-scintillating fibers calorimeter,
the first measurement of the detection efficiency for neutrons
between 20 and 180 MeV has been performed at the “The
Svedberg Laboratory” of Uppsala. The calorimeter efficiency,
integrated over the whole neutron energy spectrum, ranges
between 32-50% at the lowest trigger threshold. This result
is between three and four times larger than what expected
considering the equivalent scintillator thickness of the proto-
type. Cross-check measurement of the neutron efficiency has
been done using a NE110 scintillator. This result agrees with
previous published results.
Monte Carlo simulation shows that the origin of such
enhancement is related both to the shower-like effect due to
the inelastic processes in the calorimeter structure and to the
high sampling fraction of the detector. Preliminary comparison
between data and Monte Carlo is satisfactory. Full simulation
of the calorimeter, of the beam line and of the experimental
hall (together with local shielding) is in progress.
We performed a last data taking campaign in 2008 fall at
TSL. We used some new detectors to explain the physical
effects observed. To better describe the beam halo, we built
a new beam position monitor with (1 × 1) cm2granularity
read out using multianodes photomultipliers. We prepared a
high granularity calorimeter prototype to study the processes
inside the modules with best resolution. For this last test beam,
we also built a small calorimeter with the same structure of
the used prototype but with different sampling fraction. In
particular, this last detector has more lead in its composition,
and this is important to better study the efficiency enhancement
that is predicted by Monte Carlo has due to the secondary
interactions of neutron with the high Z material. The analysis
of the acquired data is in progress.
We thank M. Arpaia, G. Bisogni, A. Cassar` a, A. Di Virgilio,
U. Martini, A. Olivieri and all the mechanical LNF division for
the help in the construction of the NE110 reference counter
and the overall mechanical supports. We also acknowledge
A. Balla for the help in the setting of the electronic chain.
We cannot forget to thank P. Caponera and M. Rossi, which
have boldly transported all materials in the long trip from
LNF to TSL. Moreover we want to warmly acknowledge
all the TSL staff for the help during the data taking. This
work was partially supported by “Transnational access to
Research Infrastructure” TARI - The Svedberg Laboratory,
HadronPhysics I3, Contract No. RII3-CT-2004-506078.
 A. Del Guerra, NIM 135 (1976) 319-330.
 C.Birattari et al., NIM A 297 (1990), 250-257.
 C.Birattari et al., NIM A 338 (1994), 534-543.
 The KLOE collaboration, NIM A 482 (2002) 364-386.
 B. Sciascia, www.lnf.infn.it/kloe/klone/memoneueff.pdf
 LOI MARCH AMADEUS.pdf
 loi 06.pdf in http://www.lnf.infn.it/conference/nucleon05/FF.
 A.V. Prokofiev et al., PoS (FNDA2006) 016 (2006).
 A.V. Prokofiev et al., Journal of Nuclear Science and Technology,
Supplement 2 (2002), 112-115.
 A. Antonelli et al., NIM A 354 (1995), 352-363.
 A. Fass` o et al., CERN 2005-10 (2005), INFN/TC 05/11, SLAC-R-773.
 A.Fass` o et al., eConf C0303241 (2003), arXiv:hep-ph/0306267.