Test and characterization of 400 Hamamatsu R5912-MOD photomultiplier tubes for the ICARUS T600 detector

ICARUS T600 will be operated as far detector of the Short Baseline Neutrino program at Fermilab (USA), which foresees three liquid argon time projection chambers along the Booster Neutrino Beam line to search for a LSND-like sterile neutrino signal. The detector employs 360 photomultiplier tubes, Hamamatsu model R5912-MOD, suitable for cryogenic applications. A total of 400 PMTs were procured from Hamamatsu and tested at room temperature to evaluate the performance of the devices and their compliance to detect the liquid argon scintillation light in the T600 detector. Furthermore 60 units were also characterized at cryogenic temperature, in liquid argon bath, to evaluate any parameter variation which could affect the scintillation light detection. All the tested PMTs were found to comply with the requirements of ICARUS T600 and a subset of 360 specimens was selected for the final installation in the detector.


Introduction
ICARUS T600 detector is the largest Liquid Argon Time Projection Chamber (LAr-TPC) ever built for neutrino oscillation studies, successfully operated at LNGS underground laboratory (Italy) [1] It took data from 2010 to 2013 exposed to the CNGS beam from CERN and atmospheric neutrinos with a live-time factor in excess of 93% [2] After an intense refurbishing operation, carried out at CERN in the framework of the Neutrino Platform (NP01) activities, the entire apparatus was transferred to Fermilab (IL, USA). It will become the far detector of the Short Baseline Neutrino (SBN) program [3], exploiting three liquid argon detectors, placed along the Booster Neutrino Beam (BNB) line and operating at shallow depth, to investigate the possible presence of sterile neutrino states.
The realization of a new liquid argon scintillation light detection system, with high performance photomultiplier tubes (PMTs) represents a primary task of the detector refurbishing. Since the detector will be subject to a huge cosmic flux, the light detection system would allow the 3D reconstruction of events contributing to the identification of neutrino interactions in the beam spill gate.
The T600 scintillation light detection system was significantly upgraded at CERN from summer 2015 to summer 2017. The 74 outdated ETL 9357FLA PMTs mounted in the detector in 2000 [1,4] were replaced with new 360 Hamamatsu R5912-MOD units operating in liquid argon. Each PMT was tested and characterized to determine its basic functionality and to verify its compliance with the required functioning specifications. In particular all the PMTs were tested at room temperature and 60 units were also characterized at cryogenic temperature, in liquid argon bath. Measurements were carried out in different areas at CERN, where a dedicated dark-room and a cryogenic test facility were set-up.
This paper describes the test campaign carried out at CERN to characterize each PMT before installation. In particular, in section 2 the ICARUS T600 scintillation light detection system is described. The main features of the R5912-MOD are presented in section 3. Section 4 describes the employed measurement equipments and the CERN test facilities. Results are reported in section 5. A description of the base layout is also presented in Appendix A.

The ICARUS T600 scintillation light detection system
ICARUS T600 detector is made of two identical cryostats, each housing two TPCs with a common central cathode. Charged particles interacting in liquid argon produce both scintillation light and ionization electrons. These latter are drifted by a uniform electric field (E = 500 V/cm) to the anode, made by three parallel wire planes at 1.5 m from the cathode, where they are collected. Photons are detected by a PMT system mounted behind the wire planes. The information coming from both the scintillation light and the collected electrons allows to obtain a full 3D reconstruction of the interacting particle path.
The upgraded T600 light detection system consists of 360 Hamamatsu R5912-MOD PMTs deployed behind the 4 TPC wire chambers, 90 units each. Before installation the sensitive window of each PMT was coated with about 200 µg/cm 2 of Tetra-Phenyl Butadiene (TPB), a wavelength shifter to convert VUV LAr scintillation photons (λ = 128 nm) to visible light. This process was performed using a dedicated thermal evaporator and a specific deposition technique [5,6]. Mechanical supports were designed to hold up each PMT in the proper position at few millimeters behind the TPC wire planes. Each device is set inside a grounded stainless-steel cage to prevent the induction of PMT pulses on the facing wire plane. Furthermore a 50 µm optical fiber directed towards the sensitive surface will allow a timing calibration at nanosecond precision. A picture of the new light detection system is shown in figure 1.

The Hamamatsu R5912-MOD
The Hamamatsu R5912-MOD is a 10-dynode-stage PMT with an 8 in. hemispherical photocathode (see figure 2). The PMT is derived from the parent model R5912. The bialkali photocathode is deposited on a Pt underlayer in order to extend the range of the operating temperature down to about -200 • C. This PMT model was selected as a result of experimental tests carried out on different detectors samples provided by different producers, such as Hamamatsu and ETL. The main features of the Hamamatsu R5912-MOD resulting from these tests can be summarized as follows [7][8][9]: • Nominal PMT gain (G ≈ 10 7 ) and quantum efficiency (≈ 15%) suitable for the detection of ionizing particle, with energy ≤ 100 MeV; • Good uniformity of the cathodic sensitivity all over the whole PMT window (±10%);  • Linear anodic response as a function of the incident light intensity, both at room temperature and in cryogenic environment; • Pulse gain saturation above 300 photoelectrons.
A pre-series of 20 samples was delivered by Hamamatsu in 2015 for compliance testing purposes and 380 units in 2016. PMTs were selected by the producer to fulfill the performance of table 1 at room temperature, only. All the PMTs were immersed in liquid nitrogen in factory to certify their mechanical strength at cryogenic temperature.  Each device was provided with a proper base electronic circuit designed according to the voltage distribution ratio recommended by Hamamatsu. This supplies the high voltage for the grids, dynodes, and anode, and allows the signal to be picked up directly from the anode. Design and specifications are presented in Appendix A. Bases were directly welded on the PMT flying leads. Since a negative power supply is adopted, two independent coaxial cables are used to provide the PMT with the high voltage and to read out the anode signal.

Measuring equipment
All the 400 samples were tested at room temperature and 60 of them were also qualified at cryogenic temperature, in liquid argon bath, following characterization processes agreeded with the producer.
A dark-room was arranged at CERN to test in parallel up to 16 PMTs (figure 3 left). In a contiguous electronic workshop, a laser diode was set-up to produce fast light pulses at 405 nm and about 1 kHz repetition rate. The light intensity was set through calibrated optical filters mounted on two-wheel supports, allowing different attenuation combinations (1 ÷ 1000). The light was focused on the PMT windows by means of 100 µm multi-mode optical fibers. A picture of the instrumentation is shown in figure 4.
The characterization at room temperature was carried out using the electronics set-up shown in figure 5. The PMT output was integrated by means of a CANBERRA 2005 charge preamplifier  and shaped by means of an ORTEC 570 amplifier. The output distribution was recorded by means of a 8-k multichannel analyzer CAEN N915. Besides, PMT pulse shapes were acquired by means of a 10 GS/s oscilloscope (LeCroyWaveRunner 104MXI).
Tests at cryogenic temperature were carried out at CERN building 182. In order to obtain experimental conditions similar to those in the real apparatus, PMTs were directly immersed in liquid argon (T = 87 K) inside a large dewar, allowing the simultaneous bath of 10 PMTs (figure 3 right). The internal illumination was achieved by means of a single 100 µm multi-mode optical fiber. The same set-up and acquisition system described above was used together with a proper feed-through, to preserve darkness conditions and thermal insulation.

Tests and measurements
Measurements at room temperature were carried out to evaluate the performance of the devices and their conformity to the requested features. Measurements were repeated on a sample of 60 PMTs at cryogenic temperature to evaluate any parameter variation which could affect the scintillation light detection.

Signal shape
The shapes of the PMT anode pulses under single photoelectron illumination were recorded by feeding directly the anode output to the 50 Ω input of the 10 GHz bandwidth oscilloscope. Measurements were carried out with the PMT under test operating both at room and at cryogenic temperature, and at a multiplier gain G ≈ 10 7 . No significant variations were observed among the different samples or between room and cryogenic temperature, resulting in the following mean values: a leading edge of 3.9 ± 1.1 ns, a FWHM of 5.6 ± 1.1 ns and a trailing edge of 10.3 ± 1.6 ns, in good agreement with the nominal values indicated by the manufacturer.

Single electron response and gain
The response of a PMT to single photoelectrons (SER), i.e., the charge distribution of the PMT pulses integrated over the whole signal shape, is directly related to the PMT gain processes. SER studies were carried out by measuring the charge distribution of the PMT pulses induced by singleelectron excitation at different power supply values. An example of charge distribution is shown in figure 6.
Charge distributions are characterized by peak profiles which feature a good charge resolution. Spectra were fitted by means of an analytical expression which consists of an exponential distribution, which takes into account PMT and electronic noise, and 3 Gaussian functions which consider the response of the PMT to different photoelectrons. The following parameter constraints were used: X n = nX 1 σ n = σ 1 √ n A n = µ n A n−1 where X n is the position of the nth Gaussian curve with σ n width and A n amplitude. The value of the µ parameter, resulting from the fit, represents the mean number of detected photoelectrons according to Poisson statistics.
The position X 1 of the first peak allows the evaluation of the gain G. The gain dependence on the power supply applied to the PMT were measured by changing the voltage setting in a range from about G = 10 6 to about G = 5 × 10 7 . The PMT nominal voltage was defined as the power supply value to attain a gain G = 10 7 . The nominal voltage distribution for the whole set of 400 units operating at room temperature is shown in figure 7. The distribution is characterized by a quite narrow spread with mean value set at 1390 V and σ ≈ 100 V.
As an example, the results obtained at room and at cryogenic temperatures, using a single PMT, are plotted in figure 8. The signal amplitude is well represented by a power law behavior as a function of the anode voltage for both temperatures. A remarkable reduction of the gain is also evident at cryogenic temperature.
The distribution of the relative gain variation between room and cryogenic temperature, for the whole set of 60 devices tested in both conditions, is plotted in figure 9. Almost all the tested PMTs suffered a gain reduction at cryogenic temperature. Starting from a gain G = 10 7 at room temperature, the PMTs showed wide gain variations at cryogenic temperature, down to 10% of nominal value. Anyway, an increase of less than 150 V of the anode voltage was on average enough to recover the original gain factors, as shown in figure 10.
The resolution of the SER peak, or relative variance to the peak, defined as the ratio √ s = σ 1 /X 1 , where X 1 is the SER peak position and σ 1 is the width resulting from the fit, was determined for each PMT. The mean results obtained with the PMTs under test operating at a multiplier gain G = 10 7

Dark counts and noise
The response of a PMT in absence of light is represented by dark pulses and noise mainly due to ohmic leakage between electrodes on the glass and insulating surfaces of the tubes, thermoionic emission of single electrons from the cathode, field emission of electrons induced by local high electric field inside the PMT.
The dark count rate was evaluated by measuring the rate of the PMT pulses, operating in darkness condition, above the minimum value to the left of the SER peak. The distribution obtained at room temperature for the 400 tested PMTs is shown in figure 11. Dark count rate grows with the supply voltage as shown in figure 12. This behavior is well represented by a power law as already observed for the PMT gain.
The increase of dark rate when operating a PMT at cryogenic temperature is a well known effect, referred as Non-Thermal Dark Rate, and it was observed and described by many authors [10]. The precise evaluation of the dark count rate with a PMT directly immersed in liquid argon is not an easy task, due to the presence of Cherenkov radiation and scintillation produced by residual radioactive contaminants, such as Ar 39 . In order to prevent these effects, the measurements were carried out using a stainless steel chamber designed to house the PMT under test in vacuum and dark environment. The chamber was placed in a dewar that was filled with liquid argon during the cryogenic measurement. This test, which takes a quite long time for the thermalization, was carried out only on 5 PMT samples. An example of resulting rates, as a function of the discrimination threshold and for the two different operating temperatures is shown in figure 13. The curves were acquired at a multiplier gain G ≈ 10 7 . The spectrum structure is characterized by a bump profile centered in the region around one photoelectron, caused mainly by the cathode dark noise. An increase of the counting rate at cryogenic temperature, up to a factor 2, was experienced for all the tested PMTs.
The peak-to-valley P/V ratio, defined as the SER peak value divided by the minimum value to the left of the peak, is a typical indicator of the PMT capability to distinguish actual photoelectron signals from its intrinsic noise. The P/V distribution at room temperature for the 400 tested PMTs operating at G = 10 7 is shown in figure 14. The mean value, resulting from the SER distribution fitting, exceeds P/V = 3 : 1, marking very good performances of this PMT model in terms of signal/noise ratio. The variation of the P/V ratio between room and cryogenic temperature, for the 60 PMTs tested in LAr bath, is shown in figure 15. A decreases of less than 30% at cryogenic temperature, due to an increase of the PMT dynode noise at that temperature, is observed.

Photocathode uniformity
The evaluation of the photocathode uniformity required dedicated tests and was carried out for the 20 pre-series samples at room temperature, only. The PMTs were operated as photodiodes. To this purpose all the dynodes and the anode were shorted and grounded to collect all the photoelectrons emitted by the cathode which was kept at a negative voltage of about 300 V.
A collimated laser diode (Thorlabs CPS532, λ = 532 nm, 4.5 mW) was used as a continuos light source producing cathodic currents of the order of 10 µA, directly measured by a picoammeter. In order to monitor the illumination intensity and its stability, the light was splitted toward a photometer Figure 15. Comparison of the distribution of the P/V ratio at room (300 K) and at cryogenic (87 K) temperature for 60 tested PMTs. Results are referred to a power supply consistent with a gain G = 10 7 at both temperature. and the PMT which was illuminated by means of an optical fiber; a proper support was used to maintain the fiber in a fixed orientation, normal to the PMT window, while allowing to move it in various positions on the window itself. Figure 16 shows, as example, the variation of the a photocathode response as a function of the radial distance from the window center along four different axes oriented at 45 • each other. An increase of the photocathode signal approaching the window edges is recognized. This behavior was observed for all the 20 tested PMTs and it was confirmed by measurement performed by the producer.
An approximate comparison among different samples was made by averaging, for each PMT, the response data over the sensitive surface and evaluating the maximum relative variation with respect to the mean value. The resulting distribution is shown in figure 17. All the 20 PMTs show a good uniform response, within ±13% in agreement with the values indicated by the manufacturer.

Discharge and lightning
Some authors report that, at high voltage, in some PMT models it is possible to observe the presence of discharges which produce photons polluting the output signals (see for example [10]). This phenomenon is not yet well understood but it is characterized by the presence of both large light pulses ("flashes") and microscopic processes at the SER level, also in cryogenic environment.
A fast check of a possible presence of light emission was carried out during the characterization at cryogenic temperature of 60 PMTs. As described in Section 4, measurements were performed with the simultaneous immersion of 10 PMTs in LAr. This made possible to observe the behavior of each PMT both with the other 9 contiguous specimens powered ON and OFF.
Variations of the count rate between the two different operating conditions were always within the statistical errors. In addition, the presence of signal coincidences beyond the expected stochastic

Conclusions
The updated ICARUS T600 light detection system consists of 360 Hamamatsu R5912-MOD PMTs directly operating in liquid argon to reveal scintillation photons produced by ionizing particle. A total of 400 PMTs were characterized at CERN at room temperature and 60 of them at cryogenic temperature to evaluate any parameter variation which could affect the scintillation light detection.
The main obtained results can be summarized as: • SER and signal shape parameters are in good agreement with the producer nominal values, both at room and at cryogenic temperature; • The resulting gain reduction at cryogenic temperature, down to 10% of the nominal value, can be on average compensated by an increase of less than 150 V of the power supply voltage; • Dark counts and noise are in agreement with the producer nominal values at room temperature. The observed worsening at cryogenic temperature of dark counts and P/V ratio is still compatible with the correct operation of this PMT model in LAr bath; • Good uniformity of the response of the cathode was observed all over the sensitive window; • The presence of lighting phenomena were not observed; • No mechanical damages, implosions or cracking problems were observed neither during cooling/heating phases nor operating at cryogenic temperature.
All the tested PMTs were rated as compliant with the requirements of ICARUS T600 and a subset of 360 specimens were selected for the final installation in the detector.   100 Ω ± 1% R26 51 Ω ± 1% R16 0 Ω ± 1% All components are SMD, 1206 package, 200 V rated. Capacitors are C0G dielectric.