Enhancement of the DUNE FD1 X-ARAPUCA Photon Detection Efficiency and upgrade of the FD2 Photon Collector

The Photon Detection System (PDS) of the first two DUNE far detectors (FD1 and FD2) is composed of 6000 and 672 photon detection units respectively, named X-Arapuca, of different size and geometry. The PDS will complement and boost the DUNE LArTPC for the detection of non beam events: the prompt light detection will improve their tagging, and at low energies it will enable the trigger and the calorimetry of the supernova neutrinos. The X-Arapuca unit is an assembly of several components: its Photon Detection Efficiency (PDE) depends both on the design of the assembly and on the grade and the coupling of the individual components. The X-Arapuca PDE is the driver of the Photon Detection System sensitivity, that in turn determines the sensitivity of the DUNE physics reach for the detection of core-collapse supernova within the galaxy and for nucleon decay searches. In this work we present an update of the absolute PDE of the FD1 X-Arapuca baseline design, measured in laboratory: 160 units of this are deployed in the scale 1:20 FD1 prototype hosted in the NP04 cryostat at the CERN neutrino platform. Further we show how to change the baseline design of the FD1 X-Arapuca, allowing to double its PDE. Finally we review a few selected features of the photon collector of the sixteen FD2 X-Arapuca recently deployed for the FD2 scale 1:20 prototype at CERN in the NP02 cryostat, and of the last six units that integrate the latest advancements.


The Photon Detection System of the DUNE far detectors FDand FD2
DUNE (Deep Underground Neutrino Experiment) [1] is a dual-site neutrino oscillation project aiming to probe and measure the CP violating term  CP in the leptonic sector and to determine the neutrino mass ordering [2].It will also operate as an observatory for core-collapse supernova within the galaxy [3] and perform nucleon decay searches.The DUNE Far detector (FD) will consist of four 17.5 kt Liquid Argon Time Projection Chamber (LArTPC) modules, and the first two have been designed [4]: they both implement an anodic charge readout (a TPC) and a single phase photon detection system (PDS).The liquid argon emits abundant VUV scintillation light at 128 nm when excited by ionizing particles, with an absolute maximal light yield of 51000 Ph/MeV [5] corresponding to about 25000 photons/MeV in the DUNE LArTPC field at 500 V/cm.The particle energy losses populate singlet and triplet states of Ar dimers (Ar * 2 ), that de-excite with characteristic time of ≃ 6 ns (  ) and ≃ 1400 ns (  ) respectively: depending on the author and on the measurement method, the maximal   is reported to range from 1300 ns [6] up to 1600 ns [7].Both the LAr light yield and the triplet/singlet ratio depend on the energy loss mechanism; the reconstructed scintillation light profile can be used for particle identification.The light signal collected by the PDS, when combined with the charge read by the TPC, improves the calorimetry resolution, while the fast component of the scintillation light provides the trigger for non-beam events.To tag the nucleon decay events a minimum yield of >0.5 photo-electrons/MeV at the farthest position (close to the cathode) is required, corresponding to a 2.6% PDE for the X-Arapuca (XA) device.To enable supernova neutrinos calorimetry at the MeV scale the PDS requirement is an average yield of >20 photo-electrons/MeV over the detector volume, corresponding to 1.3% PDE for the XA.

Features of the FD1 and FDBaseline X-Arapuca devices
The XA device is the assembly of several components: the mechanical frame, its Viquiti® reflector lining, the SiPMs, the wavelength-shifting lightguide (WLS-LG), the glass entrance window.A cross -1 - section and scheme of the XA working principle [8] is presented in figure 1(a).The device operates a two step downshift of the LAr scintillation light: the para-therphenyl (pTP) evaporated at the device entrance window down-shifts the 128 nm to 350 nm.A fraction of the pTP photons emitted in the proper half plane (∼50%) enter the device.In the WLS-LG the 350 nm photons undergo a second downshift to ∼450 nm and only those reaching the LG surfaces at angle ( >   = 56 • ) are driven to the SiPMS located at its edges, while the dichroic filter (DF) multilayer coating deposited on the inner side of the entrance windows bounces (a fraction of) those escaping the WLS-LG ( <   ) back into the device: only the photons transmitted into the LG to the edges and those not escaping the DF may reach and are eventually collected by the SiPMS: as discussed in section 5 the DF escape probability depends on the wavelength and angle of incidence.Figure 2(a) and figure 2(b) show the X-Arapuca (XA) devices designed for the FD1 and the FD2 respectively: despite sharing the same working principle [8], their design, size, shape and readout technology differ.The active surface is ∼ 450 cm 2 (∼ 3700 cm 2 ) for FD1 (FD2), the WLS-LG/SiPM surface ratio being 26 (64).The SiPM number is 48 (160) and their mount is on rigid (flex) PCBs, the ganging is passive (active), while the bias voltages and the readout pulses are transmitted by Cu cables (optical fibers1).
The DUNE Collaboration has down-selected one SiPMs model for each of the two vendors: Fondazione Bruno Kessler (FBK) Triple-Trench (TT) [9], pixel size 50 μm, and Hamamatsu Photonics (HPK) [10] (S13360-9935) pixel size 75 μm -High Quenching Resistance (HQR).Both SiPM models are specifically designed for the cryogenic environment.Two manufacturer have also been selected for the WLS-LG: ELJEN [11] provided PVT extruded while Glass-to-Power (G2P) [12] produced PMMA as casted WLS-LGs.They have quite similar emission spectra and that of G2P is shown in figure 1(b).

The Photon Detection Efficiency of the FD1 X-Arapuca device
With two SiPM and two WLS-LG models four configurations of the XA device have been assembled: they are referred hereafter as FBK+EJ, HPK+EJ, FBK+EJ, HPK+G2P.For all of them the DF producer is OPTO [13],2 while the pTP is evaporated at the Campinas University (Brazil).In 2022, forty units of each XA configuration were deployed in the FD1 scale 1:20 prototype (pDUNE-FD1) in the NP04 cryostat at CERN.In this work we report the PDE measured on one unit of each configuration, by either the CIEMAT3 and/or the INFN Milano Bicocca (MiB)4 DUNE teams with different setups and methods.Preliminary results were released [14], and here the final results are presented with the refined analysis an determination of the key parameters.The MiB and the CIEMAT XAs, are readout by the same front-end electronics [15] and embed the same SiPMs boards: the two sets of 8 boards ×6 FBK-SIPM and 8 boards ×6 HPK-SIPM have been exchanged between the two sites, while the WLS-LG and the DFs windows are different units from the same design and the same production batch.The PDEs are measured at overvoltage (OV) values of +4.5 V and +3 V for FBK and HPK respectively, for which they both exhibit a 50% PDE at 77 K, and a similar cross-talk.Figure 3 and figure 4 (top) show the key features of the two setups: in both the VUV-LAr light is generated by 241 Am  particles (5.48 MeV).The INFN-MiB setup and method is an extension of that developed to assess the PDE [16] of a smaller XA for the SBND project [17].The sliding source located at 5.5 ± 0.1 cm distance from the XA entrance window allows to scan the device along its longitudinal axis, with a solid angle ranging 2.3 Sr-2.7 Sr from the extremes to its centre.The source is embedded in a source holder and masked to reduce the active area.The geometrical acceptance of the light cone is computed by GEANT4 based Montecarlo (MC) simulation including the source holder and the teflon mask.The device averaged PDE ( MiB ) is derived from the amplitude of the alpha peak in the pulse height spectrum.As an example a spectrum calibrated by single p.e. (SPE) response is shown in the figure 4 bottom panel: the calibration procedure is described in [14,16].The number of detected photons is then corrected for the actual LAr yield, measured at each filling i.e. for each XA configuration by the LAr triplet half-life ( t ) [18], and for the SiPMs cross-talk [19].At each filling, the  t is measured and monitored along the one-day long PDE measurement campaign.For the three XA configurations,  t was found to range from 910 ns to 1510 ns corresponding to a maximal LAr purity correction (for the 910 ns value) of ∼ 10%.As for the SiPMs X-talk, it has been measured at 77 K (11.0 ± 1.0)% and (16.05 ± 0.32)% for the HPK [19] and FBK SiPMs respectively.The PDE assessed by the MiB method ( MiB ) is the average of the PDE values measured along the XA device length (usually 6-12 points) and accounts for the devices non uniformity as the SiPM-to-WLS gaps originated from the components tolerances, (slightly) different SiPMs gain etc.The main sources of systematics are: the geometrical acceptance (7%), the SPE calibration (5%) and the actual LAr LY correction, named purity correction (2%).
The CIEMAT setup [14] consists of a 300 l cryogenic vessel with different concentric volumes, as shown in figure 3. The larger and external one (100 l) is filled with liquid nitrogen (LN 2 ) and the smaller one (18 l) is where the XA is located.The gas Argon (GAr) is liquefied there by thermal contact with the LN 2 of the surrounding volume controlling the pressure parameters to regulate the temperature and carry out the liquefaction without freezing the LAr. Figure 3 shows that the XA faces at one site the 241 Am source, the two reference HPK VUV4 SiPMs (S13370-6075CN) [10] and a Photo-Multiplier Tube (PMT) (R6836-Y00) [20] as complementary photo-sensors, all embedded in a black box.The XA entrance window is irradiated by the LAr scintillation light through the 2.3 cm diameter hole corresponding to a ∼ 1.8 Sr solid angle.As discussed in [14], the PDE is retrieved at one spot by two independent methods: the first ( MAD ) is independent from the actual LAr yield while relying on the absolute determination of the reference VUV SiPMs PDE and cross-talk and geometrical acceptance, with the associated systematics of 14% and 11% respectively.The second ( ′ MAD ), neglects the reference SiPMs and is basically as for the INFN-MiB.The final results from the two setups are summarized in table 1: the measured PDE ranges from 2.1%-2.5% for the HPK-G2P configuration and 1.3%-2.2%for the others.The results confirm the finding of [16] i.e. the G2P WLS plate shows superior LY and light transmission performances of the EJ.The XA with the HPK SiPMs has higher PDE than with the FBK.This may be related to latest findings on the TT-FBK SiPMs, showing that the external ring of pixels is passive.The results from the two independent setups and methods are very much aligned and consistent within the errors.The assessed values are in average significantly lower than those reported by previous tests on smaller XA devices [16,21] with different components configurations, and of the expected values [22].(Bottom) Two spectra collected with the source at the center (blue; 2.7 Sr irradiation solid angle) and at the bottom end (red; 2.3 Sr) of the XA device, calibrated in number of detected photoelectrons as described in [16].Reproduced from [16].© 2021 IOP Publishing Ltd and Sissa Medialab.All rights reserved.
Table 1.The PDE (  (%)) of the X-ARAPUCA for the four XA configurations deployed at the NP04 CERN facility in pDUNE-FD1, as measured with the two independent methods and setups.

Enhancement of the FD1-XA Photon Detection Efficiency
After the assessment of the PDE of the Baseline FD1 XA, efforts to improve the design of its Photon Collector design were undertaken aiming to a PDE > 3%.5Despite the constraints arising from the frozen mechanical design, we implemented few simple but relevant changes that allowed to reach a PDE of 5% as shown in figure 5(b).First, we reduced the too generous WLS-LG-to-SiPM gap present at room T, by embedding in the assembly a WLS-LG about 1.2 mm wider: the FD1-XA frame lacks of a mechanism to compensate the WLS-LG shrinking at cryo-T, therefore the ∼ 1.2 mm gap present at room-T further increases at cryo-T, since the PMMA thermal expansion coefficient is a factor ∼ 10 53% is the PDE value embedded in the DUNE physics simulations.
-5 -larger of the frame material (G10).Next we inserted Viquiti® lined blocks between the SiPM.6These two changes in joint action allowed to increase the PDE from ∼ 2.5% (black marker series) to ∼ 3.8% (red marker series) as shown in figure 5(b).The closer distance of both the SiPMs and the reflector to the lightguide and the shield of the SiPM (inactive) lateral sides increases the PDE.
A major enhancement was finally achieved by implementing a diagonal cut of about 40 • slope in the WLS-LG as shown in figure 5(a): it increased the PDE up to ∼ 5% (figure 5(b) blue series) This allows to break the inefficient paths7 and bounce the photons towards the photosensors, hence reducing the effective average optical path length.All the changes were studied by GEANT4-based light propagation simulations not shown here for space constraints.
The statistical errors are irrelevant, and the systematics are quoted in section 3. The PDEs ratio for the different XA configurations cancels the systematics.Our system shows a high stability and reproducibility: this is attested by the double markers of the red and blue series (figure 5) that represent the PDE evaluated on data taken in subsequent days of the measurement campaign.
The red and blue series are with the ZAOT [23] DF designed for the LAr environment (n = 1.24) and with cutoff at 420 nm for angle of incidence (AOI) = 45 • as discussed in section 5, while the reference black series corresponding to the baseline XA configuration is with the OPTO DF designed for a cutoff = 400 nm, AOI = 45 • and n = 1.0 (air).In a separate measurement not shown here we measured that the XA with the ZAOT DF performs about 10% better than with the OPTO ones.

Optimization of the FD2-XA Photon Collector
For the scale 1:20 prototype of the FD2 PDS, deployed at CERN in the NP02 cryostat, the XA has been newly designed: the dichroics filters (DF) are optimized for the LAr environment (n = 1.24 at 430 nm), with cutoff wavelength (  ) at 420 nm for angle of incidence (AOI) = 45 • .In order to increase the VUV active surfaces of this large-size XA device, the DF surface was increased a factor 2.3 with respect to the FD1 XA design (144 × 144 mm 2 instead of 97 × 97 mm 2 ) hence reducing the mechanical frame ribs surface.The two vendors scheme was pursued with two new industrial partners: Photon Export (PE) [24] and ZAOT [23].

Conclusions
The Photon Detection Efficiency of the DUNE FD1 X-Arapuca detector, has been assessed in LAr by two independent labs to range from 1.5% to 2.5% depending on the SiPM&WLS-LG type.This is on the lower side of the requirement for the supernova neutrino tagging and calorimetry.Efforts to enhance the PDE successfully lead to ∼ 5%, while respecting the mechanical constraints of the device design.As for the FD2 X-Arapuca unit we presented the characterization of the newly designed Dichroic Filters and of the modified WLS-LG, aiming to maximize its PDE.9 9A preliminar ∼ 2.0% PDE was reported (F.Di Capua, The Photon Detection System in the far detector module of the DUNE experiment, TIPP (2023)).

Figure 1 .
Figure 1.(a) Schematic of the XA working principle.Reproduced from [4].The Author(s).CC BY 4.0.(b) The photoluminesce (PL) spectra of the pTP (purple) and of the second downshifting chromophore embedded in the WLS-LG (blue).The dichroic filter cutoff (red dashed line) is also shown.

Figure 2 .
Figure 2. (a) Three FD1 half modules composed of two XA each: the front-end electronics boards serving two half-modules or 4 XA devices are visible at the top.(b) One FD2 XA Module equipped with its (closed) electronic box: the Dichroic Filters are the upgraded ones (see section 5).

Figure 4 .
Figure 4. (Top) The FD1 X-Arapuca device mounted on the bottom flange of the INFN-MiB LAr chamber and the GEANT4-simulated irradiation by the 241 Am source.(Bottom) Two spectra collected with the source at the center (blue; 2.7 Sr irradiation solid angle) and at the bottom end (red; 2.3 Sr) of the XA device, calibrated in number of detected photoelectrons as described in[16].Reproduced from[16].© 2021 IOP Publishing Ltd and Sissa Medialab.All rights reserved.

Figure 5 (
b) shows the impact of the proposed changes (red and blue series) on the PDE of the BL X-Arapuca design (black).

Figure 5 .
Figure 5. (a) The X-Arapuca embedding a WLS-LG with a 40 • cut and the reflector lined blocks between the SiPMs; (b) The PDE values measured with the MiB method described in section 3 as a function of the source position along the device, including the geometrical acceptance correction.
After a few test batches each vendor manufactured with its own technology the 384 DF units for the 8 Cathode (double side) + 8 Membrane (single side) XA units of the Module-0.Samples of the Photon Export vendor were measured at IFIC8 with a monochromator, and for ZAOT at UniMiB with a spectrophotometer.The DF sample are characterised in pure water (n = 1.33), the best proxy at room temperature for LAr.The Transmission curves (Tc) measured as a function of the wavelength, for several beam-to-DF AOI, are shown as an example in figure6.Both DF designs despite based on different technologies show satisfactory transmission of the pTP light (∼ 350 nm) from 0 • to 50 • and the reflecting range is actually optimized from 410 • to 500 • nm for AOI 30 • to 50 • , while for AOI > 50 • the pTP emission is largely cut out and doesn't enter the photon trap.Optical simulations later performed at IFIC indicated that the PDE can be further improved by setting   at 480 nm to extend the transmission of pTP light for AOI > 50 • .The cutoff change was implemented for the latest 80 DFs for the 6 FD2 XA units for the Module-1 prototype at CERN.As for the large area (607 × 607 mm 2 ) WLS-LG the chromophore concentration, was scaled down to 24 mg/kg, aiming to match the WLS attenuation length (figure 7(b)) at the peaking wavelength of the chromophore emission with the FD2 optical path that in the FD2-XA reaches up to ∼ 2 m.To compensate the corresponding absorbance loss at 350 nm, the thickness of the WLS-LG is increased up to 5.5-6.0 mm, covering almost completely the SiPMs surface.The lightguide is now the driving photon collection mode to the SiPMs, at the expenses of the reflection at the DF surface.The PDE simulations show an enhancement.Dedicated measurements are planned to investigate and assess it.

Figure 6 .Figure 7 .
Figure 6.The transmission curves (Tc) measured in pure water (n = 1.33) for different angle of incidence of the ZAOT DF measured with a spectrophotometer.The wavelength is on the X axis, the Tc in % units is on the y axis: the WLS-LG photoluminescence curve in a.u. is overlapped and graphed on the secondary y axis.