вход по аккаунту



код для вставкиСкачать
Nuclear Inst. and Methods in Physics Research, A 904 (2018) 81–92
Contents lists available at ScienceDirect
Nuclear Inst. and Methods in Physics Research, A
journal homepage:
The central neutron detector for CLAS12
S. Niccolai a ,∗, G. Hull a , J. Bettane a , P. Chatagnon a , B. Garillon a ,1 , B. Genolini a , B. Guegan a ,
M. Guidal a , M. Imre a , M. Josselin a , D. Marchand a , A. Maroni a , B. Mathon a , G. Murdoch b ,
T. Nguyen Trung a , J. Peyré a ,2 , J. Pouthas a ,3 , P. Rosier a ,4 , D. Sokhan b , C. Theneau a , R. Wang a
Institut de Physique Nucléaire d’Orsay, 91406 Orsay, France
University of Glasgow, Glasgow G12 8QQ, United Kingdom
Neutron detector
Plastic scintillator
Light collection
Time resolution
The Central Neutron Detector, a recently constructed scintillator barrel which is used in CLAS12 at Jefferson Lab
to detect 0.2–1 GeV neutrons at backwards angles, is here described. The motivations and R&D tests leading to its
final design, based on three radial layers of coupled paddles with one-side light readout by photomultipliers plus
‘‘U-turn’’ lightguides on the other side, are outlined. The performance of the detector, evaluated from cosmic-ray
tests and simulations, which satisfies the physics requirements, is reported.
1. The JLab 12-GeV upgrade and CLAS12
The CEBAF (Continuous Electron Beam Facility) accelerator has
delivered up to 6 GeV of high-duty-factor electron beam for hadronic
physics research to the three experimental Halls (A, B, and C) of the
Jefferson Laboratory (JLab, USA) from 1995 to 2012. More than a
hundred experiments were completed, deepening our understanding of
the strong interaction and making JLab a world-leading facility in the
experimental study of hadronic matter. An energy upgrade of CEBAF
to 12 GeV was performed in order to pursue the experimental study of
the confinement of quarks (via the search for hybrid mesons) and of
the 3-dimensional quark–gluon structure of the nucleons. The CEBAF
upgrade to 12 GeV was achieved via 5.5 recirculations through its two
pre-existing linacs, in which 10 new high-gradient cryomodules were
added. The 12-GeV beam is available for Hall D, a new experimental hall
devoted to hybrid meson studies. An 11-GeV beam is instead delivered
to the other three halls, where the capabilities of the existing detectors
were enhanced to suit the new experimental program.
The new Hall-B detector, CLAS12 (Fig. 1), is composed of two parts:
a Forward Detector (FD) and a Central Detector (CD). The FD has
similar characteristics to the old CLAS [1], with improved resolutions.
The coils of a toroidal magnet segment the detector into six azimuthal
sectors, each of which is equipped for identification of charged and
neutral particles and tracking. Its acceptance covers the polar angles
between 5◦ and 40◦ for charged particles and between 2◦ and 40◦
for photons. The CD, equipped for the identification and tracking of
backwards-recoiling charged particles, covers polar angles between 40◦
and 135◦ , with full azimuthal coverage. All of its detector components
are housed in a compact superconducting solenoid magnet, which serves
the functions of shielding the tracking detectors from electromagnetic
backgrounds and of providing the uniform magnetic field necessary for
both the momentum analysis of charged particles at large angles and
the operation of a dynamically polarized target.
2. The central neutron detector: physics case and requirements
Measuring Deeply Virtual Compton Scattering (DVCS) on a neutron
target ( → ′ ′ ) is one of the necessary steps to complete our
understanding of the structure of the nucleon in terms of Generalized
Parton Distributions (GPDs) [2–4]. DVCS on a neutron target allows one
to operate a quark-flavor decomposition of the GPDs, when combined
to results for DVCS on a proton target. Moreover, it plays a complementary role to DVCS on a transversely polarized proton target in the
determination of the GPD , the least known and least constrained GPD
that enters Ji’s sum rule [3], which links integrals of GPDs to the total
angular momentum of the quarks. To start the experimental program
of DVCS on the neutron at JLab-12-GeV, beam-spin asymmetries for
n-DVCS ( → ′ ()) will be measured with the upgraded 11-GeV
∗ Corresponding author.
E-mail address: (S. Niccolai).
Current address: Johannes Gutenberg-Universitat Mainz, 55099 Mainz, Germany.
Current address: Centre de Science Nucléaires et de Science de la Matière, 91406 Orsay, France.
Current address: Université Populaire de Caen, 14000 Caen, France.
Current address: IMPMC, 75005 Paris, France.
Received 16 January 2018; Received in revised form 9 July 2018; Accepted 10 July 2018
Available online 20 July 2018
0168-9002/© 2018 Published by Elsevier B.V.
S. Niccolai et al.
Nuclear Inst. and Methods in Physics Research, A 904 (2018) 81–92
Fig. 2. Missing mass squared of the  system, for the n-DVCS channel, simulated
with our event generator. The different colors correspond to different combinations of
particles being detected without finite-resolution smearing or with realistic resolutions.
(For interpretation of the references to color in this figure legend, the reader is referred
to the web version of this article.)
Fig. 1. The CLAS12 detector. Its forward part includes: a superconducting toroidal
magnet, high- and low-threshold Cherenkov counters (HTCC and LTCC), 3 regions of
drift chambers, forward time-of-flight counters (FTOF), electromagnetic calorimeters (EC),
and a low-angle inner calorimeter (Forward Tagger, FT, not visible) for the detection of
high-energy photons; the baseline elements of the backward part are a superconducting
solenoid magnet containing a silicon vertex tracker (SVT) and time-of-flight counters
• good neutron identification capabilities for the kinematic range of
interest (0.2 <  < 1.2 GeV/c, 40◦ <  < 80◦ );
• neutron momentum resolution  ∕ within 10%.
2.1. Constraints
CEBAF polarized-electron beam and the CLAS12 detector. The electron
and the DVCS photon will be emitted at small angles, and thus will be
detected in the forward part of CLAS12 (with the photon either in the
EC or in the FT), while the neutron will be emitted predominantly (for
∼80% of the events) at  > 40◦ in the laboratory frame, with average
momentum around 0.5 GeV/c. This points to the necessity of adding a
neutron detector (hereafter named Central Neutron Detector, or CND) to
the CD, that in its baseline design has very limited detection efficiency
for neutrons — they can be detected in the CTOF, with about 2%–3% of
The available radial space in the CD is limited by the presence of the
CTOF and of the magnet, which leave about 10 cm free. However, the
CTOF can also be used to detect neutrons, adding a couple of percent
of detection efficiency. The central tracker can be used as a veto for
charged particles. Finally, it is important to remind that there is a
surrounding magnetic field of 5 T, which complicates the choice of the
light-collection sensor.
Extensive GEANT4 simulations and R&D studies were devoted to
examine the various options for the CND and its possible photodetectors.
After considering and then rejecting the option of a ‘‘spaghetti calorimeter’’ made of lead and scintillating fibers – due to its unacceptably high
efficiency for photons with respect to neutrons – the retained design for
the detector is a barrel of standard plastic scintillator bars of trapezoidal
cross section, all with their long sides parallel to the beam direction. This
geometry is similar to the one of the CTOF.
As stated in the previous section, one of the two requirements of the
CND is good neutron identification capabilities. If the charged particles
are vetoed by the central tracker, the only particles left that can be
mistaken for neutrons are the photons. Using plastic scintillators, the
most straightforward way to distinguish neutrons from photons is by
measuring their time of flight (TOF) and compare the values of :
With the aid of the CLAS12 Fast Monte-Carlo tool (FASTMC), the
requirements in terms of angular and momentum resolutions on the
detected neutrons were determined. The kinematical variables of the
scattered electron () and of the DVCS photon (), computed by our nDVCS event generator, were ‘‘smeared’’ using the values of resolutions
provided by FASTMC. For the photon detection at low angles (2.5◦ −4.5◦ )
the Forward Tagger (FT) is used. Its energy and angular resolutions were
parametrized in the simulation according to its design specifications.
The CND requirements were determined by studying the missing
mass (‘‘’’) of the  system, which is a quantity one can cut on
to ensure exclusivity for the n-DVCS channel by minimizing the  0
contamination. First of all, without applying any resolutions on the
electron and photon kinematical variables, and varying the smearing on
the neutron kinematical variables, it was shown that the resolution on
the neutron momentum plays the major role in determining the width of
(), while the effect of the angular resolutions is less important.
Varying either  or  by a factor of 200 (from 0.1◦ to 20◦ ) increases
the width of () by about 30% more, while the same increase by a
factor of 200 (from 0.1% to 20%) on the neutron momentum resolution
 ∕ worsens the resolution of the missing mass by a factor of 40.

where  is the speed of light and  is flight path of the particle from the
target to the scintillator bar, that can be obtained, in our geometry, as
 = 2 + ℎ2 ,
where  and ℎ are the hit position along the z axis (oriented, in
our geometry, with the beam direction) and in the radial direction,
respectively. To obtain  one must measure the time of the hit at both
ends of the scintillator bar:
 = ⋅   ⋅ (  − ℎ ),
where   is the effective velocity of light propagation in the scintillator
material. To know ℎ it is necessary to have radial segmentation: ℎ will
be given by the distance between the target and the middle of the hit
Introducing the realistic resolutions on the electron and photon
calculated by FASTMC, it appears that if the neutron momentum
resolution is kept below 10% its effect is negligible with respect to the
other particles. In particular (Fig. 2, green curve), the photon resolution
is responsible of 94% of the width of the missing mass.
Therefore, considering that the detection capabilities of CLAS12 for
electrons and high-energy photons are fixed, the requirements of the
CND are:
S. Niccolai et al.
Nuclear Inst. and Methods in Physics Research, A 904 (2018) 81–92
Our GEANT4-based simulation studies [5] shown that to ensure a
good photon/neutron separation for the neutron momentum range of
the n-DVCS reaction the CND has to be equipped with photodetectors
ensuring a time resolution of about 150 ps.
3. Early R&D studies
The first part of our R&D studies was focused on exploring the timing
performance of various magnetic-field resistant photodetectors, to be
employed at the two ends of the scintillator bars, in the high-magneticfield region of the CD. Measurements of time resolution with cosmic
rays were carried out using silicon photomultipliers (SiPMs), avalanche
photo-diodes (APDs), and micro-channel-plate photomultipliers (MCPPMTs). None of these devices has however been retained. A 1 × 1 mm2
SiPM was tested, and it was rejected because, due its small active
surface, it produced a too small number of photoelectrons (∼1) and
hence yielded a too big time resolution (∼1 ns). Matrices of SiPMs
(4 × 4 elements of 3 × 3 mm2 area) have also been considered, giving
promising results in terms of time performance (∼200 ps). However,
their use would have required a very complex and costly customized
readout electronics, calling for a too long period of dedicated R&D.
The APD gave too high time resolution ( ≃ 1.4 ns), due to its big
rise time. Good timing resolution ( ≃ 130 ps) was obtained for the
MCP-PMTs, but a dramatic gain loss was observed when operating
these devices in a magnetic field. A loss of gain of a factor 104 was
observed when the field intensity passed from 0 to 5 T. Another reason
for not pursuing further the micro-channel-plate PMTs option was their
lifetime: we computed the expected flux of optical photons on the CND
photodetectors due to electromagnetic background produced over the
duration of our experiment, and it turned out to be more than an order
of magnitude higher than the limit quoted in literature [6] after which
the quantum efficiency of the MCP-PMT drops.
Fig. 3. Drawing of the Central Neutron Detector inserted in the CLAS12 solenoid.
Table 1
Dimensions (mm) of the scintillator bars of the CND. The layer numbers go from the
innermost (1) to the outermost (3). The thickness of all bars is 30 mm.
Lower base
Higher base
Fig. 4. The two set-ups adopted to test the light loss due to the use of the u-turn light
4. Description of the CND
4.1. Geometry
• using one NE102 scintillator bar read out at both ends by a
Photonis XP20D0 PMT;
• using two NE102 scintillators, still read-out by two XP20D0 PMTs,
but coupled at one end with a triangular-shaped U-turn light guide.
As none of the magnetic-field-resistant photodetectors proved to
be suited for the requirements of the CND, we pursued an alternative
solution. Our idea consists in reading the scintillation light at the
backward end of each scintillator bar, by means of a standard PMT
placed in the low-field region and connected to the scintillator via a
∼1.5-m long bent light guide. The forward end of the bar is instead
optically coupled to the neighboring paddle via a ‘‘U-turn’’ light guide
that redirects the scintillation light to the neighboring bar. The light
travels through the neighboring bar and is then collected by the PMT
connected to its end. The final detector design is a barrel, coaxial
with the beamline, made of trapezoidal scintillator bars, read out via
standard PMTs. In order to optimize the light collection by matching
the scintillator surface and the PMT entrance window the detector is
divided into 48 azimuthal segments and 3 radial layers, for a total of
24 blocks,5 144 scintillator bars, 144 PMTs, and 72 U-turn light guides
(Fig. 3). The thickness of all scintillators is 30 mm. The other dimensions
are summarized in Table 1.
For both tested prototypes, the scintillators and light guides were
wrapped in aluminized mylar for light collection optimization and with
black tape for light tightness. Two small scintillators, 1-cm thick and
with 3 × 3 cm2 of surface, coupled to two Photonis-XP2282 PMTs,
were placed above and below the NE102 scintillator bar being tested.
The coincidence signal of the two small scintillators, named ‘‘Top’’ and
‘‘Bottom’’, defined the hit position of the passing cosmic ray and was
used as trigger of the data-acquisition system. The Top and Bottom
scintillators were placed on a mechanical support that allowed a rigid
translation of the trigger system along the length of the scintillator
bar. In this way we could measure, for each configuration, the number
of photoelectrons collected on a defined PMT (‘‘Left’’) for five trigger
positions, symmetrically distributed along the scintillator’s length. The
results are shown in Fig. 5: re-directing the light to the neighboring
scintillator via the U-turn light guide causes an overall light collection
loss of only a factor 2. Furthermore, this first exploratory test, albeit
not optimized for time measurement, yield a time resolution which
was close to the experimental requirements for the detector, and thus
reassured us on the feasibility of the project.
4.2. Validation of the U-turn concept
The main concern regarding the U-turn concept was the amount of
scintillation light that is lost redirecting part of it to the neighboring
paddle. Cosmic-ray tests were carried out in order to estimate this effect.
We compared the number of collected photoelectrons as a function of
the hit position for two different configurations (Fig. 4):
4.3. CND components
Once the detector concept was defined, we performed comparative
tests to choose all its elements, always optimizing the time resolution,
A ‘‘block’’, or ‘‘sector’’, is formed by three radial layers of coupled pairs of
scintillator bars.
S. Niccolai et al.
Nuclear Inst. and Methods in Physics Research, A 904 (2018) 81–92
which is the key parameter to ensure n/ separation, and containing
costs. To this aim we realized many different prototypes and performed
comparative time-resolution and light-yield measurements to choose the
best suited scintillator, PMT, wrapping material, magnetic shielding,
shape of the U-turn light guide and glue for the optical coupling. The
outcomes of these tests are discussed in detail in the following. For all of
these tests cosmic rays were chosen as the source of irradiation, and the
external trigger system, described in Section 4.2, was adopted. The time
resolution and the light-collection efficiency of the various prototypes
were evaluated as a function of the hit position, defined by the trigger
system. This was achieved by analyzing the ADC and TDC distributions.
The ADC distributions are fitted with a Landau curve to compute the
number of photoelectrons produced at the PMT photocathode by a MIP
passing through the scintillators. The time resolution is estimated as the
 of the gaussian function fitted to the TDC spectrum.
Fig. 5. Number of photoelectrons collected on the Left PMT (‘‘L’’) as a function of
the distance from it. The blue full squares represent the number of photoelectrons
measured with the single-bar configuration, and the red open circles are the number of
photoelectrons collected on PMT Left for the setup with double bar plus U-turn light guide.
4.3.1. U-turn light-guide shape
The first set of time-resolution and light-collection measurements
was devoted to the choice of the shape for the U-turn light guide
that couples two neighboring scintillators in the CND. To this aim
two different prototypes were realized. Each prototype employed a
70-cm-long EJ200 scintillator bar and an equally sized light guide,
wrapped in aluminized mylar and black tape. Part of the scintillation
light was directly collected on a PMT (XP20D0 by Photonis, named
‘‘Direct’’ PMT) while the rest of the light was redirected, via the Uturn, to the 70-cm-long light guide. The scintillation photons had to
travel throughout all the length of the guide before being collected
by a second PMT (XP20D0 by Photonis, named ‘‘Indirect’’ PMT). The
two prototypes employed respectively a triangular and a semi-circular
U-turn light guide. The material chosen for the light guides is PMMA
(Polymethyl Methacrylate). The light collection and the time resolution
were measured for five hit positions symmetrically distributed along the
scintillator’s length: ‘‘Position 0’’ identifies the events originating in the
center of the scintillator bar; ‘‘Position 1’’ and ‘‘Position 2’’ refer to the
trigger placed towards the Direct PMT while ‘‘Position –1’’ and ‘‘Position
–2’’ refer to the trigger placed towards the U-turn.
It was found that with a semi-circular light guide around 40% more
light is collected on the Indirect PMT than using the triangular one
(Fig. 6, top). The advantage of employing the round-shaped U-turn is
also evident in the bottom plot of Fig. 6, where the time resolution,
presented as a function of the trigger position, is systematically better
than that measured with the triangular light guide. The outcome of this
first test drove our decision to employ a semi-circular light guide in the
final design of the CND.
4.3.2. Reflector material
The prototype with the semi-circular U-turn light guide was then
used to choose the reflector material for the CND. We wrapped it in turn
with aluminized mylar, VM2000 and aluminum foil, and we measured,
for each configuration, the light collection and the time resolution,
for five positions of the trigger system, as described in the previous
sections. As it is shown in the top plot of Fig. 7, using the VM2000
as reflector one collects considerably more light on the Indirect PMT,
with respect to the other wrapping materials. However, the improved
charge collection does not imply a better time resolution when using
the VM2000. Indeed, the good reflective properties of this material
lead to the collection of delayed, multiply reflected light with poor
timing properties. As the time resolutions of the prototypes equipped
with mylar and aluminum foil were comparable (bottom plot of Fig. 7),
the latter was chosen as the reflective material for the CND due to its
lower price, ease of wrapping and its high opacity, which minimizes
light ‘‘cross talk’’ between scintillator bars.
Fig. 6. Number of photoelectrons (top) and time resolution (bottom) for the semi-circular
(circles) and triangular (triangles) U-turn light guide.
scintillator bars, each 66 cm long, 3 cm thick and 3.5 cm wide, joined at
one end by the semi-circular U-turn light guide and each connected, at
the other end, to a 1.5-m-long bent light guide that directs the light to
the PMT. The Top–Bottom trigger system, described in Section 4.2, was
adopted to acquire data with cosmic rays. The prototype was equipped,
in turn, with two Hamamatsu PMTs (R2083 and R9779). While the
R2083 PMT is known to be the best performing one in terms of time
properties [7], the R9779 still provides good timing characteristics [8]
at a considerably reduced price. We also tested two Thorn EMI 9954A
PMTs provided by the CLAS collaboration and previously used in the
Large Angle Calorimeter. The time resolution measured with the onelayer prototype equipped in turn with the different PMTs is presented
in Fig. 8. The R9779- and the 9554A-based prototypes provided a
time resolution at most 10% and 30% worse, respectively, than that
obtained with the R2083-based one. In order to verify how a 10% change
in timing resolution would affect the performance of the detector in
4.3.3. Photomultiplier tube
In order to select the best suited PMT for the CND a prototype
representing one layer of the CND was built. It consists of two EJ200
S. Niccolai et al.
Nuclear Inst. and Methods in Physics Research, A 904 (2018) 81–92
Fig. 9. Results of GEANT4 simulations for the CND:  for neutrons and photons as a
function of momentum, for the two different PMTs tested. The error bars are defined
as 3, where  is the fitted gaussian width of each  peak. Blue: photons, R2083; red:
neutrons, R2083; purple: photons, R9779; green: neutrons, R9779. (For interpretation of
the references to color in this figure legend, the reader is referred to the web version of
this article.)
EJ200 [11]. The vendors report, for these scintillators, the same properties: light output around 64% of the Anthracene one, wavelength of
maximum emission at 425 nm, decay time of 2.1 ns, refractive index
of 1.58, and light-attenuation length of 380 cm. For this test, we built
two different one-layer prototypes, with the same geometry and trigger
system as described in Section 4.3.3. Both prototypes were wrapped in
aluminum foil and equipped with Hamamtsu R9779 PMTs. The time
resolution and charge collection results show essentially no differences
for the two prototypes thus confirming the equivalence between the two
tested materials. The Eljen scintillator was preferred for its considerably
reduced cost, with respect to Bicron BC408.
The trapezoidal bars, with dimensions as reported in Table 1, were
delivered by Eljen. All surfaces were polished and diamond-tool finished
at the factory.
Fig. 7. Number of photoelectrons (top) and time resolution (bottom) as a function of hit
position for the three reflector materials that were tested.
4.3.5. Magnetic shielding
The light produced in the CND scintillators reaches the photomultipliers via 1.5-m-long PMMA light guides, as the PMTs are placed outside
the region of high magnetic field. However, some stray field persists
even in this region, and the photomultipliers need to be shielded. The
average magnitude of the field at the location of the CND PMTs is 215
G, and its direction forms an angle  ∼ 107◦ with the beam direction.
Considering that the axis of the PMTs is at around 37◦ (Fig. 10), the
magnetic field is roughly perpendicular to it (at an angle ∼70◦ ). In these
conditions, a cylindrical shielding made up by a thin layer of -metal
and a thicker one (few mm) of mild steel is quoted in the literature as
the optimal solution [12].
Tests were carried out using a solenoid magnet available at the
Linear Accelerator Laboratory (LAL, France), in order to determine the
optimal thickness for the mild-steel shielding. An R9779 phototube6
was shielded with a cylindrical 1-mm-thick layer of -metal and two
different thicknesses of mild-steel cylindrical shielding: 2.5 mm and
5 mm. The -metal and mild steel cylinders are 215 and 298 mm long,
respectively, thus exceeding the PMT window of at least 75 mm, for
optimum shielding (Fig. 11). The shielded PMT was placed inside the
magnet, at its center. Light from a LED was sent onto the photocathode
via an optic fiber, and the amplitude of the output voltage of the tube
was recorded on a digital oscilloscope as a function of the value of the
magnetic field. Due to the limited space within the magnet and the
size of the phototube, the maximum relative angle between the field
direction and the PMT axis that could be tested was 30◦ . The results
Fig. 8. Timing resolution for each tested PMT as a function of hit position along the
scintillator bar. The full points represent the direct PMT, while the empty points indicate
the indirect PMT. The different symbols and colors correspond to the different PMTs under
terms of particle identification and resolutions, we ran GEANT4-based
Monte-Carlo simulations (described in detail in Section 8). The results,
presented in Fig. 9, indicate a negligible impact on the performance
of the CND, if equipped with the R9779 Hamamatsu PMTs. After the
completion of these tests, Hamamatsu started the production of a new
PMT: the R10533 [9]. It has the same detection characteristics of the
R9779, at unaltered cost, but it has two more stages of multiplication (10
dynodes instead of 8). Thanks to the higher gain, by a factor of 10, of the
R10533 with respect to the R9779, the use of an external amplification
system for the PMT output signals is not necessary. This consideration
led our choice to use the PMT Hamamatsu R10533 to equip the CND.
4.3.4. Scintillator material
In order to optimize the time resolution of the CND and considering
the allowed length for the active volume (∼70 cm), a fast scintillator
material with reduced attenuation length is needed for the detector
design. In particular we identified and tested two different scintillators
as candidates for the CND: Bicron BC408 [10] and Eljen Technology
These tests were carried out using the R9779 PMT, while we finally adopted
R10533. Still, we consider that the obtained results hold also for R10533, as it
is the same PMT as R9779, differing only for its two extra dynodes.
S. Niccolai et al.
Nuclear Inst. and Methods in Physics Research, A 904 (2018) 81–92
Table 2
Results of the direct measurement, with a Hall probe, of the magnetic field inside two kinds
of shielding, for three different relative angles () between the field and the shielding axis.

B (2.5-mm shielding)
B (5-mm shielding)
150 G
122 G
30 G
20 G
More measurements were performed later to test the effect of
perpendicular field and determine the optimal thickness of the shielding.
We increased the current of the magnet up to its maximum allowed
values and measured the magnitude of the longitudinal and transverse
components of the stray field just outside the bore using a Hall probe.
We found its values compatible with what expected at the location of
the PMTs in CLAS12. We then performed a measurement similar to the
previous one, but this time placing the PMTs horizontally just outside
the magnet bore. No matter the thickness of the shielding (2.5 or 5 mm),
we observed no change in the signal amplitude for fields ranging from
230 to 300 G, for both  = 90◦ and  = 110◦ .
A last set of measurements took place, using the same magnet,
but measuring directly the field inside the PMT shielding with a 3dimensional Hall probe, and varying the angle between the field and
the shielding. The results, for different angles between the field and
the probe, and for the two shielding thicknesses tested, are summarized
in Table 2. This measurement confirmed that adopting the 5 mm-thick
mild-steel shielding is the safest option for the CND.
Finally, a set of fine-grid simulations, using TOSCA, were performed
including the design of our shieldings, to quantify the field at the
photocathodes of our PMTs. The results of the simulations confirmed
our experimental findings, showing negligible field values, at most of
a couple of tenths of a Gauss, for both the parallel and perpendicular
components of the field with respect to the PMT axis.
Fig. 10. Side-view drawing of the Central Neutron Detector.
Fig. 11. Drawing of the PMT shielding for the CND. Both the -metal and the mild-steel
exceed the PMT length of 75 mm.
4.3.6. High voltage divider
The completely resistive high-voltage dividers of the CND were
designed following the resistors repartition, and thus the voltage distribution ratio, suggested by Hamamatsu [9]. The tube-base assembly
was developed at IPN Orsay (IPNO) with the aim to mechanically
match the mild-steel PMT shielding, for a compact and robust design.
The base bleed current is 380 μA. Considered the average number of
photoelectrons for MIPs (∼400), and the estimated rates for particles
depositing at least 1 MeV of energy in CLAS12 (∼50 Khz, obtained
from realistic simulations and consistent with the rates observed in the
commissioning of CLAS12), we foresee a mean anode current of the
order of 3 μA. This is considerably smaller than the PMT limits quoted
by Hamamatsu (0.1 mA). Moreover, the base bleed current is roughly a
factor 100 bigger than the mean anode current, as recommended by the
PMT manufacturer.
4.3.7. Light guides
The light produced in the scintillator bars of the CND reaches the
PMT photocathodes by means of light guides. In particular, for each
scintillator bar, three separated elements made of PMMA are used to
direct the scintillation light towards the photo-detector (Fig. 13). The
smaller element has one face directly glued to the scintillator while the
other, inclined with an angle of ∼70◦ , is connected to the long light
guide. The latter, in turn, is glued to a fish-tail light guide that ends with
a 42-mm-diameter circular cross section, for a proper optical coupling
with the 2-inch-diameter entrance window of the PMT (Fig. 13). The
small light guide has a trapezoidal cross section to match the scintillator
surface. The long light guide has a rectangular cross section to the side
connected to the PMMA fish-tail and was machined for ∼350 mm of
its length in order to present a trapezoidal cross section toward the
small light guide element, Fig. 13. The dimensions of the light guides
are reported in Table 3.
Fig. 12. Amplitude (in mV) of the output voltage of the R9779 PMT as a function of
the magnetic field (in Gauss), for two different orientations ( = 0◦ and  = 30◦ ) of
the phototube in the field and for two different thicknesses of the mild-steel shielding
(2.5 mm and 5 mm). The blue arrow indicates the absolute value of the field produced by
the CLAS12 solenoid at the position where the PMTs of the CND are placed.
summarized in Fig. 12, while showing little effect for a 30◦ variation of
the angle between the field and the PMT axis, point to the need of using
the thicker shielding in order to sustain a 215-G field. However, as in
our measurement the field was roughly parallel to the axis of the tube
–configuration for which the cylindrical shielding is supposed to be less
effective [12] –, these results are a worst-case scenario.
S. Niccolai et al.
Nuclear Inst. and Methods in Physics Research, A 904 (2018) 81–92
Table 3
Dimensions (mm) of the CND light guides. The layer numbers go from the innermost (1)
to the outermost (3). The ‘‘small’’ guides have trapezoidal cross section. The ‘‘long’’ guides
are trapezoidal at the scintillator side and rectangular at the PMT side. The fish tail has a
circular cross section at the PMT side.
dimensions (mm2 )
Small (1)
Small (2)
Small (3)
Long (1)
Long (2)
Long (3)
(35.92 - 39.87) ×30
(40.00 - 43.95) ×30
(44.08 - 48.03) ×30
(35.92 - 39.87) ×30
(40.00 - 43.95) ×30
(44.08 - 48.03) ×30
⊘ 42
Fig. 13. Light guides and scintillators, for one block of the Central Neutron Detector.
Table 4
Results for the tension and shear tests realized to compare the mechanical strength of
M-Bond200 and BC600.
4.3.8. Glue
For the mechanical and optical coupling of the scintillator/lightguide and light-guide/light-guide interfaces of the detector, we tested
two different glue candidates: the M-Bond200 (cyanoacrylate) [13] and
the Bicron BC600 (epoxyde) [14], providing a refractive index of 1.49
and 1.56 respectively, which match well the refracting index of PMMA
light guides and of the scintillators.
In order to choose between the two glues we used them to couple two
small pieces of PMMA together and one piece of PMMA with a BC408.
Then we evaluated the mechanical strength of the junctions performing
a tension and a shear test. For the tension test we kept the two glued
elements vertically fixed on a table and measured the load necessary
to disconnect them; for the shear test the two elements were lying
horizontally and the load was applied from one side up to the gluing
breaking point. The average results of the tests, summarized in Table 4,
clearly indicate that the M-Bond200 glue is stronger than BC600, for
both tension and shear. Another advantage of this glue is its short curing
time: while the M-Bond200 cures in few seconds, the epoxyde BC600
needs several hours for a complete polymerization. Light transmittance
measurements were also performed for the two glues, over a spectrum
between 350 and 850 nm in wavelength. Both glues had good performances over the whole spectrum, with transmittances above 80%.
BC600 provided about 5% more light yield than M-BOND. However, in
consideration of the large number of parts to be glued in the CND, the
short curing time, together with the much better mechanical strength,
led to the choice of the M-Bond200 glue.
M-Bond 200
Tension - Load [MPa]
Shear - Load [MPa]
PMMA - BC408
5. Detector assembly
The Central Neutron Detector was assembled in the mechanic shop
of IPNO. The assembly of each of its 24 blocks consisted of five different
Fig. 14. Photo of the aluminum plate used to improve the light-guide rigidity and hold
the block onto the CND mechanical support structure.
• polishing the surfaces to be glued together;
• gluing the surfaces at the scintillator/light-guide and
light-guide/light-guide interfaces;
• wrapping each scintillator/light-guide element with aluminum foil
and black tape;
• forming one layer with a pair of scintillator/light-guide elements,
gluing the u-turn;
• assembling three layers to form one block.
Completed the assembly, a 10-mm-thick aluminum plate is inserted
at the level of the fish-tail light guide to improve the mechanical strength
of the block and to allow the fastening of the block onto the CND
structure, as shown in Fig. 14.
For the block assembly and the gluing procedure we designed and
built three positioning tools that allowed to ensure ± 0.1 mm of planarity
The wrapping procedure required particular care in order to respect
the keep-in volume requirements for the CD. For the wrapping, 30μm-thick aluminum foil and 0.2-mm-thick black scotch tape were used.
Considering a total tolerance for the scintillator bars of 0.2 mm and a
total allowed imperfection thickness of 0.2 mm, it was ensured that the
total thickness of each layer be less than 30.8 mm. The thickness of each
layer was manually measured before assembling it into a block.
5.1. Mechanical structure
The support structure designed to integrate the CND into CLAS12 is
made up of six separate aluminum arches that are fastened together to
form a ring, which is in turn attached to the solenoid magnet by means
of stainless-steel brackets. The six arches are assembled together using
15-mm-thick aluminum plates to increase the rigidity and to ensure the
planarity of the assembly. Once the ring is in place, these plates are
removed and replaced with plastic ones to limit the conductivity of the
ring. For each bracket, which should support the weight of the PMTs,
S. Niccolai et al.
Nuclear Inst. and Methods in Physics Research, A 904 (2018) 81–92
Fig. 15. The Central Neutron Detector, fully assembled in a mock-up of the CLAS12
solenoid at IPNO.
of the plate and of the arch (for a total of 120 kg), we calculated a
maximum stress of 50 MPa, and a deflection equal to 0.1 mm. On the
arch, instead, the computed maximum stress is 16 MPa, producing a
deflection of 0.17 mm.
At IPNO we built a mechanical prototype to reproduce the magnet
geometry in order to test the mounting of the structure, the positioning
of the blocks as well as the stability and strength of the CND assembly.
In Fig. 15 a photo of the CND assembled at IPNO is shown.
Fig. 16. Time (in TDC channels) as a function of accumulated charge (in ADC channels)
for the 6 PMTs of a block of the CND.
6. Characterization with cosmic rays
Each block was individually tested with cosmic rays after being
assembled. The same set of six PMTs was used for all blocks. The high
voltage was set to have the same gain for all the PMTs (1.5 × 106 ).
Different trigger and DAQ systems, with respect to the top–bottom
ones described beforehand, were adopted for this systematic test of
the CND blocks. A self-triggered system was employed, recording an
event every time a track crosses the three layers and gives a signal
above threshold for all six PMTs. For the characterization of each block
we acquired 2 millions events. To identify each scintillator bar/PMT
element in the block, we named the layers H, M, and L, for the ‘‘highest’’
(the outermost), the ‘‘middle’’ and the ‘‘lowest’’ (the innermost) layer,
respectively. We also labeled ‘‘1’’ and ‘‘2’’ the scintillator bar/PMT on
the left and on the right, respectively, of a given layer. For each layer,
the signals of the two PMTs are sent to an active splitter (gain equal to
1). One of the two signals then goes to a charge-ADCs module (LeCroy
2249A — CAMAC). The other, after passing through a ConstantFraction Discriminator (GAN’ELEC FCC8 — CAMAC), is fed to a mean
timer (SEP 365) and, after an ad-hoc delay, to a CAMAC TDC (LeCroy
2228A, 50ps/ch). The signals from the three mean-timers of the three
layers are then fed to a coincidence unit (LeCroy 465) to form the trigger
signal, which is sent to the ADC gate and TDC common-start inputs. For
every event, the ADC and TDC signals are therefore recorded when a
coincidence of the three mean timers occurs.
Fig. 16 shows distributions of raw TDC versus raw ADC, obtained for
the six PMTs of the three layers of one block of the CND. It is interesting
to notice, for each PMT, the two peaks corresponding, respectively, to
the indirect light (low energy deposit and longer time) and to the direct
one (higher energy deposit and shorter time). The gap in the middle
corresponds to the U-turn.
Fig. 17. ADC spectra, for one of the six channels of a block of the CND. Left: direct light.
Middle: indirect light. Right: all hits. The red curves are fits with Landau distributions.
(higher energy) and indirect (lower energy) cases. The separation of
direct and indirect was done, event by event, based upon the TDC values.
The acquired ADC distributions are fitted with a Landau curve. The Most
Probable Values of the Landau fits provide an estimation of the number
of photoelectrons produced at the PMT photocathodes by the direct and
indirect scintillation light.
Fig. 18 shows the measured number of photoelectrons, for the direct
and indirect light, as a function of the block number, for the two PMTs
of the higher layer (i.e. H1 and H2). The distributions of photoelectrons
for the other layers are comparable to the presented one. An average
variation of the order of 10% in the amount of light directly collected on
the PMT is observed over the 144 channels of the CND. These variations
are essentially due to differencies in the light collection efficiency, which
are in turn due to the quality of the wrapping, of the gluing between the
scintillator bars and the light guides, and of the optical coupling with
the PMTs.
6.2. Effective velocity
6.1. Number of photoelectrons
The effective velocity of light transmission in the scintillator bars was
calculated dividing the length of the bars by their ‘‘time length’’, given
by the width of the TDC distribution. The light velocity distributions for
L1 and L2, are shown in Fig. 19, as an example. The overall variation
of   , for the different blocks and for the different scintillators in a
given block, is estimated to be less than 4%. As expected, the measured
In Fig. 17 the ADC spectra, for one of the six channels of one
block (i.e. H1 of block # 6) of the CND, are displayed. The right-most
plot is the ADC distribution after subtraction of the pedestal. The left
and middle plots are the ADC distributions for, respectively, the direct
S. Niccolai et al.
Nuclear Inst. and Methods in Physics Research, A 904 (2018) 81–92
Fig. 18. Measured number of photoelectrons as a function of the block number, for the
scintillation light detected by the H1 and H2 PMTs. The full and the empty symbols
represent, respectively, the direct and indirect light.
Fig. 21. Time resolution for the 24 blocks of the CND, measured with cosmic rays in
triple-coincidence mode. The dashed line indicates the target performance (150 ps). An
8% systematic uncertainty is assigned to all measurements.
the fact that in our value the effects of the light guides and the u-turn
are folded in.
6.4. Time resolution
Fig. 19. Values of the effective velocity for the L1 and L2 scintillators, for the 24 blocks
of the CND.
To define an average time resolution for our setup in the three-layerscoincidence trigger configuration, we use the method adopted for the
CLAS TOF system [15] and inspired by the work done in [16]. Let us
consider the stack of three layers of scintillators under test, which we
label as , , and , going from top to bottom. Let us also define 
and  the two sides at which the timing is read. The resolution of the
reference system (formed by the  and  scintillators) is determined by
 +  −  − 
The spread in  , denoted by  , is the time resolution of each PMT,
assuming all four PMTs are equal. Designating the timing for a cosmicray particle passing through counter S as
then the quantity
Fig. 20. Number of photoelectrons versus distance from the PMT, for the six PMTs of a
block of the CND. The exponential fits yield the light-attenuation length.
 +  +  + 
is the time a cosmic ray passed through scintillator  relative to the
reference system. The width of the distribution,  , is corrected by
the resolution of the reference system,  , to obtain the resolution of .
Thus, with minimum-ionizing particles, the resolution of the  counter
is calculated as the standard deviation

)2 .
light velocity in the scintillators does not exceed the nominal value
  = ∕ = 19 cm/ns, calculated for a refractive index  = 1.58.
6.3. Attenuation length
Fig. 20 shows the number of photoelectrons, obtained from the
ADC spectra, for the six PMTs of a given block (# 6), as a function
of the position of the hits along the scintillator bars. The hit position is
defined dividing the TDC spectra, for the direct light, into five equally
spaced intervals. For each time bin, then, we plot the corresponding
ADC spectrum and fit it with a Landau distribution to evaluate the
corresponding number of photoelectrons. From the exponential fits
of the distributions of number of photoelectrons versus position, we
estimate the attenuation length, which is an essential parameter for the
energy calibration of the detector. The average attenuation length for
the whole detector is 149 cm, with an overall standard deviation of
25 cm. This differs from the nominal value for EJ200 (380 cm) due to
With this method, we obtained an average  = 148 ps over the 24
blocks. The resolutions of the 24 blocks are shown in Fig. 21. The
systematic uncertainty, represented by the error bars on Fig. 21, has
been estimated by repeating various times the same kinds of data takings
as well as by using subsets of the same data set, and it is of the order of
7. Electronics
In order to operate the PMTs, high voltages (typically in the range
of 1500 V) are provided to them by multi-channel CAEN SY527 power
S. Niccolai et al.
Nuclear Inst. and Methods in Physics Research, A 904 (2018) 81–92
supplies. The HV boards adopted for the CND are CAEN A734N (16
channels, 3 kV max voltage, 3 mA max current). The signal of each
PMT is sent to an active splitter. The three splitter modules used for
the CND were originally developed by IPNO for the G0 experiment
(Hall C, JLab) [17]. Each module is an active 64-channels splitter with
unity gain, so that there is no loss of amplitude. The 64 SMA inputs
are placed in the back panel. In the front panel there are 8 8-channels
output connectors (DMCH) for the time signals and 4 16-channels output
connectors (FASTBUS) for the charge signals. The charge signal from the
splitter is sent to the flash-ADC 250 VXS, 16 channels/board, made and
owned by JLab. The time signal from the splitter is sent to a constant
fraction discriminator (CFD) GANELEC FCC8, originally developed for
the TAPS detector in Mainz. Each CFD module is an 8-channels CAMAC
unit with LEMO 00 input connectors and 2 × 8-pin output connectors in
differential ECL. The threshold can be set for each channel individually
using a manual switch or remote control, and no walk adjustment is
required for the module. The discriminated time signal then goes to the
TDC (CAEN VX1290A, 32 channels/board, 25 ps/channel resolution). In
total, the read-out system includes 3 splitter modules, 19 CFD modules,
5 TDC boards, and 8 ADC boards.
Fig. 22. Efficiency for the detection of neutrons having 0.4 GeV/c of momentum, as a
function of the threshold on the deposited energy. The efficiency is shown for 3 different
values of , between 50◦ and 70◦ .
8. Simulation
In order to study the performance of the CND, its geometry was
added to the CLAS12 GEANT4-based simulation package, GEMC. The
‘‘FTPT_BERT’’ physics list was adopted [18]. The timing resolution
and the energy loss due to the U-turn geometry were included in the
simulation using the values measured in the cosmic-ray tests. The main
steps of the neutron reconstruction are the following:
• a hit in the CND is considered as a neutral hit (neutron or photon)
if no hits recorded in the Central Tracker match its position7 ;
• among all reconstructed neutral hits in the CND, only those passing
a certain threshold on the energy deposition are kept;
• among these surviving hits, only those for which the hit position
–reconstructed from the Direct and Indirect PMTs timings –is
within the length of the paddle are kept;
•  is calculated for each selected hit, to exclude photons;
• the angles  and  and the momentum are computed (see Section
8.2) for the identified neutrons.
Fig. 23. Efficiency for the detection of neutrons emitted at 60◦ , as a function of
momentum, for 7 different values of the threshold on the deposited energy, from 1 to
5 MeV.
Simulations, which included all the components of the CD, were run to
evaluate the efficiency of the CND for neutrons, its ability to discriminate between neutrons and photons, and its angular and momentum
resolutions. Neutrons and photons of momenta varying between 0.1 and
1 GeV/c and having polar angles  varying between 50◦ and 70◦ were
generated at fixed azimuthal angle ( = 3.75◦ ). The results obtained
with these simulations are described in Sections 8.1–8.3.
with increasing threshold, ranges between 12% at the lowest thresholds
and 7% at the highest ones. This can also be seen in Fig. 23, where
the efficiency for neutrons emitted at 60◦ is plotted as a function of
momentum, for various values of the threshold on the energy deposition.
Fig. 24 shows instead the efficiency as a function of the momentum of
the neutron, at a fixed energy threshold of 3 MeV, and for different
values of . All of these plots were done with a cut rejecting hits with
time of flight larger than 8 ns. This cut has been applied to suppress
the events in which the neutrons interact in the magnet (without
depositing energy in the CND) and rescatter or produce secondary
particles hitting the CND at a later time, compromising the PID and the
determination of the angles. This cut, along with a choice of threshold on
the reconstructed deposited energy of a few MeV (3 MeV is the chosen
value at the present stage), is effective in removing these secondary hits.
8.1. Efficiency
The detection efficiency is defined here as the ratio between the
number of events for which a ‘‘good hit’’8 was reconstructed either in
the correct azimuthal bin of the CND or in any of its two neighbors
(left and right) and the total number of neutrons generated. Several
values of energy thresholds, between 1 and 5 MeV, were studied. Fig. 22
shows the efficiency as a function of the threshold, for neutrons with
momentum of 0.4 GeV/c. The different colors correspond to 3 different
values of the neutron polar angle, . The efficiency, which decreases
8.2. Angular and momentum resolutions
The resolutions on the polar angle  of the neutron that can be
obtained with the CND are strongly linked to its TOF resolution. The
angle  is in fact given by
In order to compensate for possible inefficiencies of the Central Tracker,
additional, tracker-independent constraints can be imposed, such as requiring
only ‘‘straight tracks’’ –i.e. neutrals, which are not bent in the magnetic field
–using the positions measured directly by the CND and/or the CTOF.
A ‘‘good hit’’ must, first of all, produce a signal in each of the two coupled
paddles; then, the timings of the two signals must differ enough to be compatible
with one being the direct and the other the indirect signal.
 = (180∕) ⋅ arccos(


where  and  both depend on the time measurement. Using the value
 = 0.24 ns MeV−1∕2 , deduced from the cosmic-rays measurements, for
S. Niccolai et al.
Nuclear Inst. and Methods in Physics Research, A 904 (2018) 81–92
Fig. 25.  distributions for neutrons with  = 0.2 GeV/c (black),  = 0.4 GeV/c (red),
 = 0.7 GeV/c (green),  = 1 GeV/c (blue), and photons with  = 1 GeV (purple). The
threshold on the deposited energy is 3 MeV. The plots show all reconstructed particles,
integrated over . Equal neutron and photon yields are assumed here. (For interpretation
of the references to color in this figure legend, the reader is referred to the web version
of this article.)
Fig. 24. Efficiency for the detection of neutrons, as a function of neutron momentum, for
a 3-MeV threshold on the deposited energy. The efficiency is shown for three different
values of , between 50◦ and 70◦ .
the gaussian smearing on the timing
 = √


the  resolution was studied with GEMC, as a function of neutron
momentum and  itself. The reconstructed  was defined as the average
over all layers, whenever more than one layer had a good hit. The
angular resolution  , obtained via gaussian fits of the simulated 
distributions, increases slightly with the angle and also is fairly constant
as a function of neutron momentum, and its value is between 1.8◦ and
3◦ .
The resolution on the azimuthal angle is directly connected to the
total number of scintillator bars along . The angular size of each bar,
, being 7.5◦ ,  is given by ∕2 = 3.75◦ .
The resolution on the neutron momentum, which is obtained knowing  and having performed the particle identification, according to the
= √
1 − 2
Fig. 26.  versus momentum for neutrons (black) and photons (green) with momenta
between 0.2 and 1 GeV. The error bars are defined as 3, where  is the fitted width of
each  peak. The threshold on the deposited energy is 3 MeV. (For interpretation of the
references to color in this figure legend, the reader is referred to the web version of this
is also strictly connected to the time resolution. The momentum resolution  ∕ ranges between 4.5% and 6%, for increasing neutron
momentum. No appreciable variations of momentum resolution are
observed by varying the neutron polar angle. As for the  reconstruction,
also in this case the reconstructed momentum is computed as the
average over all layers, whenever more than one layer had a good hit.
Fig. 25 shows the comparison between the  distributions obtained
for neutrons of various momenta (0.2, 0.4, 0.7 and 1 GeV/c) and for
1-GeV photons. All particles in this plot are emitted at  = 60◦ . A
small portion of the neutrons having momentum of 1 GeV/c can be
taken as photons, as their  distributions begin to overlap, while the
n/ separation is clear for lower momenta — which correspond to most
of the range of interest for n-DVCS, as only about 8% of the events are
expected to have  > 0.9 GeV/c. This is also evident from Fig. 26, where
the error bars correspond to 3, where  is the gaussian width of each
 distribution. Equal neutrons and photon yields are assumed for this
study. This assumption was justified by simulating the expected photon
rates for both physical photons and electromagnetic background.
8.3. Particle identification
Since the charged particles passing through the CND will be vetoed
by the Central Tracker, the only particles that could be mistaken for
neutrons in the CND are the photons. The efficiency of the CND for
photons has been estimated by simulations, and it is slightly bigger
than the one for neutrons (of the order of 15%, having little energy
dependence). Neutrons can be discriminated from photons by means
of their . Therefore, the  distributions that can be obtained with the
CND for neutrons and photons were studied with the help of the GEMC
simulation. After choosing a good hit,  is computed as

 = ℎ2 + 2 .
8.4. Estimated radiation dose
In order to evaluate the effects of the radiation induced by electromagnetic backgrounds on the CND, GEANT4-based simulations were
run, simulating the interaction of the beam electrons with the target
material and with the whole CD at the nominal CLAS12 luminosity.
From our simulations we learned that running with an 11-GeV electron
beam at a luminosity of 1035 cm2 s−1 produces a total deposited
energy per layer between 3.7 Gy/year (for the outermost layer) to
ℎ is the distance from the vertex to the middle of the layer where the
hit took place,   is the reconstructed time of flight and  is the
reconstructed position of the hit along the scintillator bar.
S. Niccolai et al.
Nuclear Inst. and Methods in Physics Research, A 904 (2018) 81–92
4.5 Gy/year (for the innermost layer). For the scintillator material of
the CND (Polyvinyl toluene) it is reported that a 23% degradation
in the light transmission is observed starting from doses of 105 Gy.
Degradation of the PMMA constituting the light guides and of the glue
connecting scintillators and guides are reported to happen for doses
around, respectively, 104 Gy and 103 Gy. The simulated dose per year
being at least 3 orders of magnitude below these values, we should not
observe any damage to the CND blocks over the 15 or more years of
running foreseen for CLAS12. We have also estimated that no damage
will incur to our photomultipliers due to the radiation dose. According
to our simulations, in fact, the total charge accumulated at the anode
will be about 4 orders of magnitude less than the typical values (500 C)
for which the PMTs are quoted to reach their half life time.
in other nDVCS experiments [20], and whenever the detection of the
recoil neutron may be required ( ∗ program, for instance, or all the
deeply-virtual meson production reactions on the neutron).
This work was supported by IN2P3-CNRS (France) and European
(Sixth Framework Program I3-HP) funds.
[1] B.A. Mecking, et al., Nucl. Instrum. Methods A 503 (2003) 513.
[2] D. Müller, D. Robaschik, B. Geyer, F.-M. Dittes, J. Horejsi, Fortschr. Phys. 42 (1994)
[3] X. Ji, Phys. Rev. Lett. 78 (1997) 610; Phys. Rev. D 55 (1997) 7114.
[4] A.V. Radyushkin, Phys. Lett. B 380 (1996) 417; Phys. Rev. D 56 (1997) 5524.
[5] S. Niccolai, et al., JLab Letter of Intent LOI12-09-001.
[6] N. Kishimoto, et al., Nucl. Instrum. Methods A 564 (2006) 204–211.
[12] Photomultiplier tubes - Basics and applications, Hamamatsu Handbook, 2006.
[15] E. Smith, et al., Nucl. Instrum. Methods A 432 (1999) 265.
[16] R.T. Giles, et al., Nucl. Instrum. Methods A 252 (1986) 41.
[17] D. Androic, et al., Nucl. Instrum. Methods A 646 (2011) 59.
[19] S. Niccolai, et al., Jefferson Lab experiment E12-11-003.
[20] S. Niccolai, et al., Jefferson Lab experiment E12-06-109a.
9. Conclusions
This article presented the requirements, the design, the R&D, and
the performance of the Central Neutron Detector for CLAS12. It consists
of a barrel of three layers of scintillators coupled at their front ends
with u-turn light guides and read out at their back sides by ordinary
PMTs connected to the bars via 1-m-long bent light guides and placed
in the low-field region of the CND. The CND was installed in the CLAS12
solenoid, and subsequently started its data taking, in the fall of 2017.
Our GEANT4-based simulations, calibrated with measurements carried
out with the CND in cosmic rays, show that the efficiencies obtainable
with this detector and its photon-rejection capabilities will be sufficient
to collect good statistics on the beam-spin asymmetry for the neutronDVCS reaction over a wide phase space, using the allocated beam time
for CLAS12 with deuterium target [19]. This detector will also be used
Без категории
Размер файла
2 817 Кб
niman, 2018, 029
Пожаловаться на содержимое документа