Prototype Tests and Construction of the Hadron
Blind Detector for the PHENIX Experiment
C.Woody, W.Anderson, B.Azmoun, C.-Y. Chi, A.Drees, A.Dubey, M.Durham, Z.Fraenkel, J.Harder,
T.Hemmick, R.Hutter, B.Jacak, J.Kamin, A.Kozlov, A.Milov, M.Naglis, P.O’Connor, R.P.Pisani,
V.Radeka, I.Ravinovich, T.Sakaguchi, D.Sharma, L.Shekhtman, A.Sickles, S.Stoll, I.Tserruya, B.Yu
Abstract— A Hadron Blind Detector (HBD) has been
constructed as part of the detector upgrade program for the
PHENIX experiment at RHIC. The HBD is a proximity focused
windowless Cherenkov detector operated with pure CF4 that will
be used to detect single and double electrons in relativistic heavy
ion collisions and provide additional rejection power against
Dalitz pairs and photon conversions. The detector consists of a 50
cm long radiator directly coupled to a set of triple GEM detectors
equipped with CsI photocathodes to detect UV photons produced
by electrons emitting Cherenkov light. A full scale prototype of
the HBD was built and tested in order to study its performance
under beam conditions. Tests with the prototype demonstrated
good separation between electrons and hadrons using pulse height
discrimination and cluster size. The final detector has now been
constructed and installed in PHENIX and is presently undergoing
commissioning in preparation for its first round of data taking
during the next heavy ion run at RHIC. Results of the beam test
of the prototype, as well as on the construction and initial testing
of the final detector, are presented in this paper.
Hadron Blind Detector (HBD) has been constructed as
part of the upgrade program for the PHENIX
Experiment at the Relativistic Heavy Ion Collider (RHIC) at
BNL . The HBD will allow the measurement of electron-
positron pairs from the decay of the light vector mesons (ρ, ω
and φ) and the low-mass pair continuum in Au-Au collisions at
energies up to √sNN = 200 GeV.
The device is a windowless Cherenkov detector using pure
CF4 as a radiator in a proximity focus configuration utilizing a
set of triple GEM detectors equipped with cesium iodide (CsI)
photocathodes. The photocathode is evaporated on the top
Manuscript submitted on November 19, 2006. This work was supported in
part by the U.S. Department of Energy under Prime Contract No. DE-AC02-
A.Dubey, Z.Fraenkel, A.Kozlov, M.Naglis, D.Sharma, I.Ravinovich, L.
Shekhtman and I.Tserruya, are with the Department of Particle Physics,
Weizmann Institute of Science, Rehovot, Israel
B.Azmoun, A.Milov, R.P.Pisani, T.Sakaguchi, A.Sickles, S.Stoll and
C.Woody* are with the Physics Department at Brookhaven National
Laboratory, Upton, NY (*Contact author: E-mail: woody @bnl.gov).
J.Harder, P.O’Connor, V.Radeka and B.Yu are with the Instrumentation
Division at Brookhaven National Laboratory, Upton, NY
W.Anderson, A.Drees, M.Durham, T,Hemmick, R.Hutter, B.Jacak, and
J.Kamin are with the Physics Department, Stony Brook University, Stony
C.-Y. Chi is with Nevis Labs, Columbia University, Irvington, NY.
surface of the uppermost GEM which produces photoelectrons
that are amplified with a gain of ~ 5x103 by the GEM detector.
The combination of a windowless detector with the CsI
photocathode and CF4 results in a very large bandwidth (from
6 to 11.5 eV) and a very high figure of merit (N0 ~ 840 cm-1).
One expects approximately 36 detected photoelectrons with a
50 cm long radiator, which ensures a high level of single
electron efficiency and double hit recognition. Electrons
traversing the radiator produce Cherenkov light in the form of
a “blob” on a pad readout plane with a pad size approximately
equal to the blob size (~10cm2) resulting in a low granularity
A significant R&D program was carried out to develop the
various components of this detector that has now been
completed [2,3]. A prototype detector was constructed to test
the key parameters of the device and to measure its
performance under actual test beam conditions. The final
detector has now been constructed and installed in PHENIX,
and is undergoing commissioning in preparation for its first
round of data taking during the upcoming run at RHIC.
II. THE HBD CONCEPT
Figure 1 shows the configuration of the triple GEM
detectors used in the HBD. The CsI photocathode is deposited
on the upper surface of the top GEM and a bias voltage is
applied between the top GEM and the mesh. Depending on the
direction of the bias field, charge produced by ionizing
particles in the upper gap can either be collected by the GEM
(FB = Forward Bias), or by the mesh (RB = Reverse Bias). In
either configuration, photoelectrons produced on the
photocathode are collected with good efficiency into the GEM
due the strong electric field near the holes. Only a very small
amount of ionization charge produced very near the
photocathode (within ~ 100 µm) is collected by the GEM. The
Forward Bias mode is therefore sensitive to hadrons and other
charged particles, while the Reverse Bias mode is essentially
sensitive only to Cherenkov light produced by electrons (hence
the term “Hadron Blind”). Numerous tests were carried out in
the lab to study the effect of suppressing the charge collected
in negative bias mode and how the reverse bias voltage effects
the photoelectron collection efficiency. These tests are
described in detail in Refs  and .
Fig. 1 Configuration of GEM detectors and CsI photocathodes in the HBD.
With a higher negative voltage on the mesh than on the top GEM (Forward
Bias), electrons produced in the top gap are collected by the GEM. With a
lower negative voltage on the mesh than on the top GEM, ionization electrons
are collected by the mesh (Negative Bias), and the detector is “Hadron Blind”.
III. DETECTOR CONSTRUCTION
The HBD vessel was designed and built at the Weizmann
Institute of Science in Rehovot, Israel. It consists of a thin
honeycomb structure for the outer vessel, which constitutes
about 3% of a radiation length of material inside the fiducial
acceptance of PHENIX, including the gas radiator. The
detector is divided into two halves, one of which is shown in
Fig. 2. Each half contains twelve triple GEM modules, each
consisting of one gold plated GEM on the top and two
standard GEMs below. The dimension of each GEM is
approximately 23 x 27 cm2. All GEMs were produced at
CERN. Out of a total of 133 foils produced, 65 standard
GEMs and 47 gold plated GEMs passed all quality assurance
tests. Forty eight standard GEMs and 24 gold plated GEMs
were used to construct the final detector.
Fig. 2 Exploded view of one of the HBD vessels.
B. GEM modules
Each GEM was tested individually and then matched in
groups of three in order to give the lowest possible gain
variation for all modules in the detector. The resulting gain
spread from module to module varied from 5-20%. Each
module was then mapped in order to study the gain variation
within a module. The gain was measured with an 55Fe source
which produced a rate of ~ 8 KHz. The source was initially
placed over one pad and the gain was measured as a function
of time in the same location for approximately half an hour.
During this time, the gain was observed to increase by
anywhere between a few percent to up to a factor of two
depending on the module before, reaching a steady plateau.
This effect is shown in Fig. 3 for a module which exhibited a
gain increase of about a factor of 1.5. This behavior is typical
of many GEM detectors and is believed to be due to an initial
charging up effect due to polarization of the polyimide foils
when the high voltage is initially applied.
Fig. 3 Gain variation of a triple GEM module during gain mapping. The
initial rise is typical of most GEM detectors and varied between a few percent
and a factor of two depending on the module.
The source was then moved to measure all the other pads
with the module and then returned to its original position
where the gain was measured again. The gain was typically
higher at the beginning of this second measurement, but then
decreased to the original plateau established during the initial
charge up period. We also found that the rates of gain increase
and decrease were somewhat different (the discharge rate was
approximately a factor of two shorter than the charging up
rate), and the charging up rate was slightly rate dependent (~
10-30% for rates between 10Hz – 8 KHz). However, once the
operating plateau was reached, the gain appeared to be stable.
C. Photocathode production
The photocathodes for the HBD were produced at Stony
Brook University in a “Clean Tent” that allowed both
fabrication of the photocathodes using a high vacuum
evaporator system, as well as assembly of the detector under
very clean and dry conditions (typically Class 10-100). Figure
4 shows the Clean Tent at Stony Brook, which contained the
evaporator seen in the back, a large, high quality glove box, a
laminar flow hood, a high vacuum storage container for the
GEMs, and numerous other auxiliary pieces of equipment.
Fig. 4 The “Clean Tent” at Stony Brook University where the CsI
photocathodes were produced and the final detector was assembled. The high
vacuum evaporator is located at the back of the tent, with the glove box on the
right and the laminar flow table on the left in the back.
The evaporator was originally constructed by the INFN and
the Instituto Superiore di Sanita in Rome, Italy and is presently
on loan to Stony Brook University . This is a very
sophisticated and high quality apparatus that is capable of
producing four HBD photocathodes at a time, along with
several small “chicklets” that were used for monitoring the
photocathode quality. GEM foils are mounted into an open a
aluminum box and placed inside the evaporator which
deposits a thin layer of CsI on the top surface of the GEM. The
rate of deposition was ~ 20 A/sec and the total thickness of the
final layer was between 2400-4500 A. The vacuum inside the
evaporator was typically 10-7 torr during evaporation. The
quantum efficiency of the photocathode was then measured in
situ inside the evaporator over the entire area of the GEM
using a remote controlled movable UV light source and current
monitor. The quantum efficiency was measured over the entire
area of the photocathode at several wavelengths from 165-200
nm, while the test “chicklets” were measured over the
wavelength range from 120-200 nm in a separate
monochrometer. Fig. 5 shows a position scan for a typical
photocathode, which shows good uniformity across nearly the
entire foil. Fig. 6 shows the quantum efficiency measured for
one of the chicklets compared to a good quality photocathode
produced at the Weizmann Institute. In general, all of the
photocathodes produced were of excellent quality and showed
very good uniformity.
Fig. 5 Spatial uniformity of several typical photocathodes produced in the
Stony Brook evaporator.
Fig. 6 Quantum efficiency as a function of wavelength of a photocathode
produced in the evaporator at Stony Brook compared to a known good
photocathode produced at the Weizmann Institute. Measurements made at
Stony Brook were done in vacuum. Measurements at the Weizmann were done
in vacuum and CF4.
The photocathodes were sealed inside a gas tight aluminum
box while still inside the evaporator and then transferred to the
glove box. The photocathodes were therefore never exposed to
air at any time during this process. The glove box atmosphere
of nitrogen was kept extremely dry and oxygen free (O2 < 5
ppm and H2O < 10 ppm). The photocathodes were installed
into the detector and electrically tested along with all of the
other GEMs. Fig. 7 shows the first half of the HBD with all of
its photocathodes installed while still in the glove box. After
all of the photocathodes and GEM modules were installed, the
detector was sealed and removed from the glove box where it
was put under its own gas flow of argon or nitrogen.
Once outside the glove box, circuit boards containing the
readout electronics were installed on the back side of the
vessel. These boards contained the preamplifiers that are
connected to the readout pads inside the detector via a series of
short, individual wires that pass through the honeycomb wall
of the vessel and are soldered to the input pads of the preamps.
The preamps were designed by the Instrumentation Division at
Brookhaven (IO-1195) and produce a differential signal in the
range from 0 to ± 1V that is delivered to a receiver and front
end module designed by Nevis labs. The front end module
contains a 12 bit, 65 MHz ADC for each channel which
digitizes the signal and sends the data via an optical G-Link to
the PHENIX data acquisition system.
Fig. 7 One half of the final HBD detector with all CsI photocathodes and
GEM modules installed inside the glove box used for assembly.
IV. FINAL DETECTOR INSTALLATION
The final detector was installed into the PHENIX experiment
during the fall of this year. Figure 8 shows the west detector
which was installed in early September. The two square panels
on the front face of the detector are heater foils that are
described in the next section. The east detector was installed
approximately one month later, surrounding the beam pipe.
Fig. 8 HBD west detector installed in PHENIX
V. HIGH PURITY GAS SYSTEM
It is extremely important to maintain low water and oxygen
levels in the operating gas of the final detector. Both water and
oxygen have absorption bands in the deep UV that would
absorb Cherenkov light and reduce the overall photoelectron
yield . Every 10 ppm of either oxygen or water results in a
loss of approximately 1 photoelectron due to absorption in the
50 cm gas radiator. In addition, water will adversely affect the
photocathode performance and reduce its quantum efficiency.
In order to maintain the highest purity operating gas in the
final detector, the HBD gas system in PHENIX has been
constructed using all stainless steel tubing and valves, and with
components which are free of silicon, grease or any materials
that are prone to out gassing. Both water and oxygen are
monitored continuously for the common input gas, and
independently for the output gas from each half of the
While the input gas can be kept very pure, out gassing of
water can occur from inside the detector due to the release of
trapped water from the GEM foils, the kapton readout plane
and the FR4 frames of the vessel. External heater foils were
installed on the outside surfaces of the vessel which can be
used to raise the temperature inside to ~40 deg C in order to
drive out as much water as possible in a conditioning mode
before the detector is actually operated for physics data taking.
This process has been very successful, and the present water
levels are down to ~10-15 ppm in the final installation and are
expected to further improve before data taking begins.
Fig. 9 Gas transmission monitor system partially constructed. From left to
right: D2 lamp and scanning monochrometer, beam splitter box, reference
PMT, gas test cells (one input, two output) with one PMT mounted..
In addition to monitoring the water and oxygen levels, we
have constructed a gas transmission monitor that will measure
the UV transmission of the input and output gas from both
halves of the detector. This will not only serve to verify the
absorbance caused by oxygen and water, but will also identify
any other possible contaminants in the gas which could cause
absorption of the Cherenkov light. The system consists of a
scanning VUV monochrometer (McPherson 234/302 with D2
lamp) with a beam splitter box containing a movable mirror
that directs a beam of light down three independent gas test
cells (one input, two output). The light impinges on
photomultiplier tubes (Hamamatsu R6835) with CsI
photocathodes operated in photodiode mode and the
photocurrent is measured using a Keithley picoammeter. Fig 9
shows the partially assembled spectrometer which will be
completed and installed in PHENIX within the next few
VI. PROTOTYPE TESTS
A full scale prototype of the HBD was built and tested
under beam conditions (200 GeV pp collisions) in the
PHENIX experiment at RHIC during the spring and early
summer of this year. The detector showed stable operation
with pure CF4 as the operating gas and worked well in the
RHIC environment. The detector also demonstrated the
expected performance in terms of “hadron blindness”. Figure
10 shows the pulse height distribution for minimum ionizing
particles obtained with the detector in forward bias (FB) mode
and reverse bias (RB) modes. The forward bias distribution
shows a clear minimum ionizing peak and is well fit with a
Landau distribution. The reverse bias mode shows the strong
suppression of the direct ionization signal, as expected.
Fig. 10 Pulse height distribution for minimum ionizing particles obtained
with the full scale HBD prototype operating in the PHENIX experiment at
RHIC. The forward bias distribution is well fit with a Landau distribution,
while the reverse bias distribution shows the expected strong suppression of
the direct ionization signal.
Figure 11 shows the pulse height distribution and cluster
size for electrons and hadrons obtained with the prototype
detector under reverse bias conditions. The electrons and
hadrons were both well identified using other detectors in
PHENIX. The electron distribution is well separated from the
hadron signal and gives approximately a 15:1 rejection factor
with an electron efficiency of 90% using a simple pulse height
cut alone. The cluster size distribution shows a much larger
average cluster size for electrons, and will be used to further
increase the rejection power in the final analysis
Fig. 11. Pulse height distribution (a) and cluster size distribution (b) for
identified electrons and hadrons with the HBD prototype operated in reverse
bias mode in PHENIX during the last RHIC run.
A novel new Hadron Blind Detector has been constructed
for the PHENIX experiment that will greatly enhance the
capability to measure low mass electron pairs at RHIC. A test
with a full scale prototype detector was carried out during the
spring of this year and demonstrated the hadron blindness
feature of the HBD concept under actual running conditions.
The construction and assembly of the final detector was
completed during this past summer, and both halves of the
final detector have now been installed in PHENIX and are
undergoing preparations and commissioning for the upcoming
run at RHIC.
 B.Azmoun et.al., “Conceptual Design Report on a HBD Upgrade for the
PHENIX Detector”, March 2005,
 A.Kozolov et.al.., “Development of a Triple GEM UV Photon Detector
Operated in Pure CF4 for the PHENIX Experiment”, Nucl. Inst. Meth.
A523 (2004) 345.
 Z.Frankel et.al., “A Hadron Blind Detector for the PHENIX Experiment
at RHIC”, Nucl. Inst. Meth. A546 (2005) 466-480.
 Evaporation facility provided by the INFN and Instituto Superiore di
Sanatia in Rome, Italy, courtesy of Prof. F. Girabaldi and collaborators at
 I.Tserruya et.al., “A Hadron Blind Detector for the PHENIX Experiment
at RHIC”, Conference Record Proceedings of the 2004 IEEE Nuclear
Science Symposium and Medical Imaging Conference, Oct 16-22, 4004,