Measurement of the absolute efficiency of the X-ARAPUCA photon detector for the DUNE Far Detector 1

Eur. Phys. J. C (2024) 84:1004 https://doi.org/10.1140/epjc/s10052-024-13393-2 Regular Article - Experimental Physics Measurement of the absolute efficiency of the X-ARAPUCA photon detector for the DUNE Far Detector 1 R. Álvarez-Garrote1, C. Brizzolari2,3, A. Canto1,E.Calvo 1, C. M. Cattadori2, C. Cuesta1, A. de la Torre Rojo1, I. Gil-Botella1,C.Gotti 2, D. Guffanti2,3, A. A. Machado5, S. Manthey Corchado1, I. Martín1,C.Massari 2,3, L. Meazza2,3,a, C. Palomares1, L. Pérez-Molina1,b, E. Segreto4, F. Terranova2,3, A. Verdugo de Osa1, H. Vieira de Souza7, D. Warner6 1CIEMAT, Avda. Complutense 40, 28040 Madrid, Spain 2Dipartimento di Fisica “Giuseppe Occhialini”, Universitá degli Studi di Milano-Bicocca, Piazza della Scienza 3, 20126 Milan, Italy 3INFN Sezione di Milano-Bicocca, Piazza della Scienza 3, 20126 Milan, Italy 4Dipartimento di Fisica, Universitá degli Studi Federico II, 80126 Naples, Italy 5Instituto de Física “Gleb Wataghin”, UNICAMP, Campinas, SP 13083-859, Brazil 6CSU, Limelight Ave, Castle Rock 80109, USA 7Laboratoire Astroparticule et Cosmologie, Rue Alice Domon et Léonie Duquet 10, 75013 Paris, France Received: 23 May 2024 / Accepted: 20 September 2024 / Published online: 7 October 2024 © The Author(s) 2024 Abstract The DUNE far detector has been designed to detect photons and electrons generated by the charged products of the interaction of neutrinos with a massive liquid argon (LAr) target. The photon detection system (PDS) of the first DUNE far detector (FD1) is composed of 6000 photon detection units, named X-ARAPUCA. The detection of the prompt light pulse generated by the particle energy release in LAr will complement and boost the DUNE LAr Time Projection Chamber. It will improve the nonbeam events tagging and enable at low energies the trigger and the calorimetry of the supernova neutrinos. The XARAPUCA is an assembly of several components. Its photon detection efficiency (PDE) depends on the design of the assembly, on the grade of the individual components and on their coupling. The X-ARAPUCA PDE is one of the leading parameters for the PDS sensitivity, that in turn determines the sensitivity of the DUNE for the detection of core-collapse supernova within the galaxy and for nucleon decay searches. In this work we present the final assessment of the absolute PDE of the FD1 X-ARAPUCA baseline design, measured in two laboratories with independent methods and setups. Preliminary results were reported in Palomares (JINST 18(02):C02064, https://doi.org/10.1088/ 1748-0221/18/02/C02064, 2023). One hundred sixty units of these X-ARAPUCA devices have been deployed in the NP04 facility at the CERN Neutrino Platform, the 1:20 scale ae-mail: [email protected] be-mail: [email protected] (corresponding author) FD1 prototype, and will be operated during the year 2024. The assessed value of the PDE is a key parameter both in the NP04 and in the DUNE analysis and reconstruction studies. 1 Introduction The Deep Underground Neutrino Experiment (DUNE) [2] is a dual-site experiment that aims to measure the neutrino oscillation parameters with a precision and sensitivity that will allow to test the CP violation in the leptonic sector and determine the neutrino mass ordering [3]. It will also perform nucleon decay and beyond standard model searches and it will contribute to the detection of astrophysical neutrinos from the galaxy, the Sun and core-collapse supernova within the galaxy [4]. The DUNE Far detector (FD) will consist of four 17 kt LArTPC modules. This liquid argon (LAr) technology will make possible to reconstruct neutrino interactions with image-like precision. The design of the four identically sized modules is sufficiently flexible for staging construction and evolving the LArTPC technology. The first FD Module (FD1) will use the horizontal drift (HD) technology, in which ionization charges drift horizontally in the LAr under the influence of an electric field towards a vertical anode, where they are read out. Four 3.5 m drift volumes are created between five alternating anode and cathode walls, each wall having dimensions of (58 ×12) m2, and installed inside a cryostat. A schematic of the TPC is shown in Fig.1. 123 1004 Page 2 of 14 Eur. Phys. J. C (2024) 84 :1004 Fig. 1 10 kt DUNE far detector module, showing the alternating 58.2 m long, 12.0 m high anode and cathode planes, as well as the field cage that surrounds the drift regions between the anode (APA) and cathode plane (CPA) [2] LAr produces abundant VUV scintillation light, emitting 51,000 photons/MeV when excited by minimum ionizing particles in the absence of a drift field [5,6]. The particle’s energy losses populate singlet and triplet states of Ar dimers (Ar∗ 2), that de-excite with characteristic times of 6 ns and about 1.6 µs respectively, emitting 127 nm photons. The high light yield and its efficient detection will enhance DUNE detector capabilities. To fulfill the supernovae neutrino program and for efficiently tagging nucleon decays, DUNE requires an average light yield of >20 photoelectrons (PE)/MeV with a minimum >0.5PE/MeV,which corresponds to a collection efficiency of 1.3% and 2.6%, respectively [7]. The photon detector system of FD1 consists of light collectormodulesplacedintheinactivespacebetweentheinnermost wire planes of the anode planes. These large-area light collectors convert incident scintillation photons into photons in the visible range that end up in photo-sensors. After investigating many light collector modules, DUNE has successfully reached a design for a light trap that can be deployed as either single-face or dual-face readout, the so-called X-ARAPUCA. In order to validate the DUNE technology, prototypes are being designed, built and tested at CERN, the ProtoDUNEs, LArTPC prototypes of 0.77kt total mass. ProtoDUNE-HD will test the final design of FD1 components at scale 1:1 in 2024. In particular, the X-ARAPUCAs will be tested for the first time there. 2 The X-ARAPUCA device The X-ARAPUCA (XA) is an improvement of the original concept of photon trapping inside a highly reflective box while using a wavelength shifting (WLS) bar to increase the probability of collecting trapped photons onto a SiPM array. The ARAPUCA technology has been validated in the run-1 of the ProtoDUNE detector, that showed its superior performances in the photon detection efficiency compared to dip-coated shift lightguides (single or double) [8]. In the XA the entrance window is a glass coated on the external side with a layer of p-Terphenyl (pTP) (∼500µg/cm2)to convert the incident 127 nm scintillation light into isotropically re-emitted 350 nm photons with an efficiency larger than 95%. The re-emitted light reaches the WLS bar where it is further downshifted to the visible range (430 nm), as illustrated in the schematic of Fig.2a. The WLS re-emitted light can be trapped by total internal reflection (θ>56◦) or escape the WLS bar (θ<56◦) and be reflected by a dichroic filter and by a reflective surface. The filter is a multilayer thin-film coating deposited onto the inner side of the entrance window, whose cutoff is at 400 nm (Fig.2b). These trapped photons propagate towards the edges of the module where the silicon photo-multiplier (SiPM) photo-sensors are placed. Simulations of the X-ARAPUCA device [9]showed the effectiveness of photon trapping inside the light guide and its dependence on factors such as light guide bulk transmittance, surface reflectiveness and optical coupling with the SiPMs. In DUNE, two different variants will be deployed: a double sided optical window variant, to equip the middle anode plane which collects scintillation photons from both the central drift volumes and a single sided variant to equip the two external anode planes. In this paper the single sided variant has been tested, in which the side opposite to the optical window is an opaque backplane lined inside with an extended specular reflector (ESR). Two different models for the WLS have been tested, the EJ-286 manufactured by Eljen Technology [10] and the other custom designed and manufactured by glass to power (G2P) [11] in collaboration with INFN [12]. Figure2b shows the emission spectra of both the G2PWLSandtheprimarywavelengthshifter.Earlierstudies conducted at INFN showed the improved PDE performances of a smaller (or half-size) X-ARAPUCA device, embedding the G2P WLS than with the EJ286 [13]. The SiPM arrays are facing two of the sides of the WLS bar perpendicular to the entrance window. An image taken during the mounting process of an XA installed in Neutrino Platform 04 (NP04) is showed in Fig.2c. The SiPMs are evenly spaced and an ESR is placed between them on the left empty space on the mounting boards, so that photons reaching the edge of the module not hitting a SiPM are reflected back into the optical module. TheSiPM activeareais 6x6mm whilethe lightguide is 4mm thick so that 1/3 of the area is exposed to the LAr allowing the detection of photons trapped by the dichroic filters. The module dimensions are 2m long and 12cm wide and it consists of 4 XA as shown in Fig.3a, also known as Supercells.InFig.3b we can see that each XA is composed of 6 dichroic filters, 1 WLS bar and 48 SiPMs grouped in 8 PCB boards (6 SiPMs each); the bias and signals are routed 123 Eur. Phys. J. C (2024) 84 :1004 Page 3 of 14 1004 Fig. 2 (a) Schematic of the XA working principle [14]. (b)The dichroic filter cutoff (red dashed line), the pTP (purple line) and the G2P (blue line) emission spectra. (c) Image taken while mounting an FD1 XA of dimensions (50 ×12)cm2 to the front end electronics by signal leading boards custom designed by INFN Sezione di Milano [15]. Four configurations of these XAs depending on the SiPMs and WLS bar type will be tested in ProtoDUNE. Two different models of SiPMs, from Fondazione Bruno Kessler (FBK) and Hamamatsu Photonics K.K (HPK), will be employed. The particular models are: FBK Triple-Trench (TT) [16], which pixel sizeis about 50 μm, and HPK (S13360-6075HSHRQ) 75 μm High Quenching Resistance (HQR) [17] both with a total effective area of about 36mm2and specifically designed for being used at cryogenics temperatures (CT). ProtoDUNE-HD also makes use of two different WLS bars, the EJ-286PS and the bar custom designed by G2P. For all the four configurations OPTO [18] dichroic filters were used. In Table 1we summarize the four combinations installed in ProtoDUNE-HD that have been characterized in this work. 3 Methodology and instrumentation Quantifying the absolute efficiency is fundamental for DUNE, as this parameter is needed to fully characterize the PDS. The PDE must be assessed at the operative conditions, at CT and with 127 nm photons. For this reason, we submerged the XA in LAr together with a low-activity electrodeposited241Amalphasource.Twoexperimentalsetupshave been established to carry out this measurement, at CIEMAT (Madrid, Spain) and at Milano-Bicocca University (Milan, Fig. 3 (a) PDS module including 4 XA cells. (b) Schematic of the assembly process with the different components [14] Table 1 The four XA configurations installed in ProtoDUNE-HD SiPMs WLS bar Testing site (A) FBK TT EJ-286PS-1 CIEMAT + MiB (B) FBK TT G2P-FB165A MiB (C) HPK 75HQR EJ-286PS-1 CIEMAT (D) HPK 75HQR G2P-FB165A CIEMAT + MiB Italy). To measure the absolute efficiency of the XA, the number of photons arriving at its surface need to be known. Two different methods have been considered: comparing the amountof lightcollected bythe XA withthe lightdetected by a calibrated photosensor, and estimating the light in the XA from the α-source energy and the known number of scintillation photons per MeV produced in LAr once the solid angle sustained by the source is determined. Both setups liquefy high purity argon gas (GAr) inside a cryogenic vessel. Thedataprocessingandacquisition(DAQ) systemconfiguration, involving the software and the electronics, is similar in both setups and a diagram is shown in Fig.4 The 48 SiPMs are passively grouped at the input of a transimpedance amplifier located at the top of the XA, the socalled cold-amplifier. The amplifier is based on a SiGe bipolar transistor (Infineon BFP640) followed by a fully differential operational amplifier (Texas Instruments THS4531), and is designed for low noise at low power, giving a voltage 123 1004 Page 4 of 14 Eur. Phys. J. C (2024) 84 :1004 Fig. 4 Schematic of the data acquisition system white noise density of 0.37 nV/√Hz at 2.4 mW per channel. To reduce the noise that may be induced in signal cables, the output of the cold-amplifier is differential. At room temperature, the so-called warm-electronics converts the differential signal to single ended and introduces a second amplification factor. The readout scheme replicates the one planned for the DUNE FD1, where the differential to single ended conversion is performed in AC using a transformer, as described in [19]. This gives an undershoot on the tail of the signals, which will need to be considered in the analysis. Both the cold and the warm electronics are the same as in the MiB and CIEMAT setups, while the adopted digitizer differs, as reported in the following Sects. 3.1 and 3.2. 3.1 CIEMAT setup The CIEMAT neutrino group made use of a 300l cryogenic vessel with different concentric volumes, whose schematic is shown in Fig.5. The larger and external one (100l) is filled with liquid nitrogen (LN2) and the smaller one (18l) contains the GAr and it is where the XA is located. In this 18l container, the GAr is liquefied by thermal contact with the LN2of the surrounding volume. This is achieved by controlling pressure parameters and regulating the temperature values necessary to carry out the liquefaction. The system is designed to perform the automatic filling of the 100l vessel with LN2from a 400l tank which is at an over-pressure of 4 bar. At the end, GAr grade 6.0 is liquefied with LN2at 2.7 bar. To reduce contamination from material outgassing, we perform successive vacuum cycles in the vessel before introducing the optical and electrical components that will be used to perform the measurements. A last vacuum cycle is done once all elements are in place. Each XA configuration is tested in data taking campaigns lasting 3–4 days. The XA is introduced in the inner vessel together with two reference HPK VUV4 SiPMs (S13370-6075CN) [17] and a photo-multiplier tube (PMT) (R6836-Y00) [20] as complementary photo-sensors. The VUV4 SiPMs are designed to have a high sensitivity for VUV light and stable performance at CT, making possible the detection of LAr scintillaFig. 5 Setup scheme used for obtaining the absolute efficiency of the XA. (a) Cryogenic vessel with its concentric volumes where GAr is liquified. (b) Diagram of the black box holding the sensors together with the 241Am source tion light. The PMT is not specifically designed to measure at CT, although it is also sensitive to VUV light, allowing us to study the scintillation light profile easily and monitor the LAr purity. Together with the reference detectors a 241Am source is held with an opaque box as shown in Fig.5.Itemitsαparticleswith5.485MeV(84.45%)and5.443MeV(13.23%) energies with an activity of (54.53 ±0.82) Bq [21]. The particles deposit their energy inside the 4cm sized black box, and the produced photons reach the XA through a hole (∅= 23mm). We ensure that no other photons are being detected by covering the rest of the XA with a black sheet. Onthe other faces ofthe box,we placethe twoVUV4 SiPMs, the PMT and a diffuser connected to a fiber. The optical fiber 123 Eur. Phys. J. C (2024) 84 :1004 Page 5 of 14 1004 Table 2 Relative distances from the α-source to the sensors and sizes of the box’s holes XA Ref. SiPM Effective area (mm2) 415.47 36.00 Distance to the source (mm) 29.0 ±0.7 26.9 ±0.3 provides light from LEDs and lasers to check the response of the system. In Table 2we summarize the detectors’ dimensions and positions in the box. The VUV4 SiPMs were calibrated at room temperature by the manufacturer; however, several studies [22–24]have shown a decrease of about 50% in the PDE at CT and a dependence with the incident angle. Considering the results ofthesestudiesandthemeasurementscarriedoutinCIEMAT labs [25]with these sensors exposed to VUV light at different angles, we will assume a PDE of (11.17 ±1.3)% at 127 nm and, 87 K and 4 V overvoltage (OV) for the model S133706075CN of VUV4 SiPMs from HPK. The cross-talk probability computed in [25] for the reference SiPMs at CT is PXT =(14.84 ±0.24)% which is in perfect agreement with the result presented in [22]. The XA signal provided by the warm-electronics is digitized by the ADC (model CAEN DT5725S [26]). The signals of the two reference SiPMs are amplified and digitized by the same CAEN module. The trigger is done at ADC level and is provided by the signal in coincidence of the two reference SiPMs. The final output consists of 20 μs waveforms with 4 ns sampling. A pre-trigger of 2 μs allows the determination of the baseline on a event-by-event basis. For calibration runs, a pulse generator provides an external trigger and synchronously pulses the light source (laser or LED). Three-μs waveforms are recorded in the calibration runs with the same time sampling. 3.2 INFN Milano Bicocca setup The INFN Milano-Bicocca (MiB) setup is an extension of the one used for the PDE measurements of the XA device adopted by the SBND project [27]. The work [13] allowed us to precisely assess the superior performance of the SBNDXAdevice equippedwith thecustomproducedPMMA based WLS [11], that is now adopted as the baseline component for boththeDUNE FD1andFD2 PhotonDetectionsystems. The setup, the procedures and the method are described in detail in [13]. Figure 6shows the XA located at the center of the stainless steel ∼25l cylindrical chamber of 250mm diameter and 550mm height. The closed chamber is located in an open 70l dewar. Thechamber is firstoutgassed downto O(10−4)mbar and then connected to a bottle of GAr grade 6.0. The open dewar is filled with LAr and the GAr liquefaction process Fig. 6 (a) The scheme of the INFN-MiB setup for the longitudinal scanning and PDE determination of the XA. (b) Picture before closing the chamber starts inside the chamber. Both the GAr flow and its liquefaction rate in the chamber are sustained by the regulation of the bottle pressure reducer. The exposed 241Am α-source (3.7 kBq) is mounted on the tip of a magnetic manipulator (rototraslator) and allowed to slide on a vertical rail, facing the XA at the distance of (55±1) mm. This allows to scan the PDEoftheXAalongitsz-axisandtomonitorwithhighprecision the LAr level inside the chamber during the whole filling process, as the alpha light pulse amplitude greatly increases when the source is in LAr. The target operational condition is reached when the XA and the front end readout circuit are submerged in LAr. Due tothe lackof GAr,in fewruns wecouldn’treach thetopof the device, hence only the data from the scanning positions fully submerged by LAr are considered in the PDE data analysis. The digitizer (CAEN DT5725 250 MS/sec 14 bits) is selftriggered by setting a threshold, that provides a trigger rate of about 1 kHz for the α-particles and 100 Hz for the muons runs respectively. 123 1004 Page 6 of 14 Eur. Phys. J. C (2024) 84 :1004 Fig. 7 Example of events generated for (a) CIEMAT and (b) MilanoBicocca setups 3.3 Monte Carlo simulations To properly assess the PDE of the XA prototype, a dedicated GEANT4 [28] Monte Carlo (MC) simulation was developed for each setup. For both of them, the scintillation photons generated by the alpha particle energy loss in LAr are emitted uniformly and isotropically, and the MC provides the number of photons reaching the XA acceptance window. For the CIEMAT setup the reference photosensors solid angles are also determined by the simulations and for both MiB and CIEMAT the geometrical acceptances uncertainties are computed by varying both the position and dimensions of the setup elements within their precision errors. Figure 7a shows the simulation of the CIEMAT setup with the dimensions presented in Table 2. The two reference VUV4 SiPMs (red), the PMT (blue) and the XA (green) are configured as sensitive materials to retrieve the number of detected photons depending on their position with respect to the source. The surrounding black box is designed with a black plastic material that fully absorbs the photons. Each event of the alpha source is simulated to have the number of photons in a random position within the sensitive area. We have determined a systematic error due to the geometrical acceptance uncertainty of 10.8%. Figure 7b shows a side view of the XA long edge side of the Milano Bicocca MC geometry. About 500 photons are drawn for visualization purposes. The 241Am deposited surface exposed to LAr is embedded in a source holder whose shape and size is included in the MC model to determine the effective light cone. The geometrical acceptance is then driven by the device-to-source distance, that is measured (5.5±0.1)cm, along the whole rail length. The uncertainty on the geometrical acceptance accounts for ∼7% and repFig. 8 (a) Persistence histogram of selected waveforms together with theaveragewaveformofonesinglePE.(b)Chargehistogramofselected peaks,the gainisdefinedbythedistancebetweenthe first andthesecond peak resents the major systematic error for the MiB method. The LAr optical properties are simulated too but for optical path in LAr of (O(10 cm)) both the absorbance and the Rayleigh scattering are negligible. 4 Data analysis 4.1 Calibration To measure the PDE, the light pulse generated in LAr by 241Am alpha particles and collected by the photosensors must be calibrated by the integrated charge of the single photon electron, i.e. by the gain factor of the SiPMs. The gain depends on the operative temperature and bias voltage, and can be affected by the fatigue effect, like for the photo123 Eur. Phys. J. C (2024) 84 :1004 Page 7 of 14 1004 Table 3 Experimentally measured SNR of the XA SiPMs at a given overvoltage (OV) and corresponding photon detection efficiency (PDE) OV PDE SNR (a) FBK TT 3.5 40 2.40 ±0.08 4.5 45 3.42 ±0.07 7.0 50 3.76 ±0.05 (b) HPK HQR75 2.0 40 2.90 ±0.03 2.5 45 3.55 ±0.02 3.0 50 4.24 ±0.02 multiplier tubes [29]. In both the CIEMAT and the MiB setup the calibrations are performed with low intensity blue light emitting sources. Typical calibration waveforms for the XA are displayed in Fig.8a. The gain is determined from the integrated charge distribution, as shown as an example in Fig.8b and defined as Gain =(μ2−μ1)whereμnisthemeanvalueoftheGaussian corresponding to nphotoelectrons. The good signal-to-noise ratio(SNR)allowsthe identificationofthepeaks corresponding to 1, 2, 3 to N PE and then the gain determination. The SNR qualifies the capability to detect a single PE (SPE) over the system noise: it depends both on the XA electronics and on the setup related disturbances. We define SNR =μ1−μ0 σ2 0+σ2 1 from the integrated charge distribution where σnis the Gaussian width of the nth peak (0 is the baseline noise peak, 1 is the 1PE peak). For each photosensor type, Table 3reports the SNR measured for three overvoltage bias values. In all cases a SNR >2 was measured, hence in all the measurements the SPE detection capability is verified. Figure 9features the measured gain versus over-voltage of XAs equipped with different SiPM models. The gain is a characteristic of the SiPM and independent of the rest of the XA elements. Dedicated cross-talk studies have been performed by the PDS Consortium for the two SiPMs models deployed in the XAs.Themeasured cross-talk probabilities(PXT, probability to have a second pixel activation after a true photo electron) are presented in Table 4for three OV bias values and are used to compute the correction for the efficiency. The cross-talk correction factors (fXT) used for the analysis are computed as follows: fXT =1 1+PXT ±PXT (1+PXT)2.(1) To asses the XA PDE and compare the performances of the different configurations, we choose the bias OV value of 4.5 OV for FBK TT and 3.0 OV for HPK HQR75, since for Fig. 9 Gainversusbiasvoltage.Dashedlinesrepresentlinearfitsof the experimental points. Unit conversion to get adimensional gain values was made using the amplifier’s gain and the electron charge Table 4 Experimentally measured cross-talk probabilities at CT OV PDE PXT (%) (a) FBK TT 3.5 40 12.68 ±0.27 4.5 45 16.05 ±0.32 7.0 50 32.47 ±0.47 (b) HPK HQR75 from [30] 2.0 40 6.6 ±0.7 2.5 45 9.0 ±1.0 3.0 50 11.0 ±1.0 these values the two models exhibit similar gain (see Fig.9) while keeping the PXT <20%. Theaveragewaveforms forthe SPEare computedfor each XAconfiguration andtheresults areshown inFig.10 forboth MIB and CIEMAT setups. The front-end electronics provides bipolar signals with a characteristic zero crossing time of 1.0 µs. Due to this behaviour we integrate only the positive part of the signal. The LAr light emission time profile has a triplet component withcharacteristic emissiontimeτt∼1.6µs>1.0µs,soto assess the integral of the charge lost in the negative lobe two methods are considered: either the alpha/muon waveforms are deconvolved (CIEMAT) by the SPE response, or the scintillation time profile is convoluted with the SPE response (MiB). The fraction of positive lobe charge is corrected after this computation. 123 1004 Page 8 of 14 Eur. Phys. J. C (2024) 84 :1004 Fig. 10 Single photo-electron normalized response obtained during calibration in CIEMAT and Milano-Bicocca setups. XAs with (a)FBK and (b) HPK SiPMs 4.2 Milano Bicocca analysis In the method adopted at MiB and used as secondary method byCIEMAT,theXAefficiency(MiB(XA))iscomputedfrom the ratio of the detected (#PE(XA)) to the expected (#Ph) light: MiB(XA)=#PE(XA) #Ph ·fcorr.(2) where #Ph is: #Ph =LYLAr Eα=35700 ph/MeV ·5.48 MeV ·, (3) The maximal LAr light yield for alpha particles in LAr LYLAr =(35700 ±2157)photons/MeV including the αquenching factor qα=(0.70 ±0.04)is from [31,32], while the geometrical acceptance () is determined by the Monte Carlo simulations discussed in Sect. 3.3. #PE(XA) is derived from the fit of the full energy peak of the calibrated alpha spectra as described later in this section. The alpha spectra are the histograms of the alpha waveformsselectedbypulseshapediscriminationcriteria (PSD)[13] and charge integrated over 1000 ns. Figure11ashowsthe capability of the MiB setup and method for particle identification by PSD when cutting on the fraction of the prompt (charge integral <600 ns) over the total (charge integral <1000 ns), named hereafter fprompt. Muons and alphas are clearly separated in the (fprompt) vs total charge plane: alphas have fprompt >0.7 and muons fprompt <0.7. More details on the alpha spectra analysis are provided later on in this section. The correction factor (fcorr) is the product fcorr =fXT ·fint ·fpurity.(4) and takes into account the cross-talk (fXT), the fraction of light falling in the waveform positive lobe (fint) and the effective LAr light yield (fpurity), the latter being related [33]to thelight quenchingimpurity (e.g.N2)concentration thatmay vary at each filling of the experimental chamber. The correction for the effective LAr yield (LYeff)isrelevant for the absolute PDE determination and to fairly compare the PDE of the different XA configurations. Up to several ppm values the impurities affect only the triplet (or slow) component of the LAr emission that for alpha particles accounts for only 23% of the LYLAr. The re-normalization factor (fpurity) accounts for the fraction of the actual triplet (τexp) to the maximal (τpure) component [33] f−1 purity =Aslow τexp τpure +Afast,(5) whereAfast =0.77andAslow =0.23aretheliteraturesinglet and triplet contributions for alphas and τpure =1600 ns is the triplet time constant for the maximal LAr LY [13,32]. At each filling of the experimental chamber i.e. for each of the tested XA configurations, the τexp is extracted from the muon waveforms analysis. A muon run is taken with the source located at the top position to limit the number of alpha eventstriggeringtheDAQandmuonswaveformsareselected by PSD criteria: the selected muon waveforms are individually deconvoluted by the SPE waveform template and the resulting normalized muon average waveform is fitted by a two exponential function convoluted with a Gaussian, providing τexp as shown in Fig.11b. The τexp, hence the quality of the LAr, is monitored and found to be stable along the alpha data taking time. The τexp measured ranges are reported in Table 5for each XA configuration together with the the corresponding fint ·fpurity. 123 Eur. Phys. J. C (2024) 84 :1004 Page 9 of 14 1004 Fig. 11 (a) Fraction of prompt over total integrated charge for an alpha run; alpha and muons populations are clearly distinguishable by cutting on the prompt fraction. (b) The deconvolved muon waveform measured with the MiB setup Table 5 The LAr triplet decay time constant τexp ranges derived from the muons analysis and the corresponding combined correction factor for the three measurements with the MiB setup, computed for the best fit in the range Measurement τexp [ns]fint ·fpurity (A) FBK +Eljen 910–1113 0.843 (B) FBK +G2P 910–1115 0.843 (D) HPK +G2P 1407–1507 0.853 Fig. 12 Computation of the charge integration correction factor for τexp =963 ns. From top to bottom: simulated time profile, time profile convolved with electronics response, fraction of integrated charge (y axis) as a function of the integration time (x axis). fint =92.77% is found for a 1000 ns integration window The waveform bipolar shape requires to assess the actual charge integrated within the positive lobe of the alpha waveforms (fint), that is in turn anti-correlated to fpurity and τexp; in fact at the increase of the latter and since it is greater than the waveform zero crossing time, a larger fraction of the late photons are lost since they fall into the waveform negative lobe. Therefore the fint ·fpurity is numerically computed as follows and described in Fig.12: the alpha particle scintillation time profile for a given τexp, (top panel), is convoluted withtheSPEtemplateand providestheexpectedpulseshown in the middle panel. The integration correction factor (fint)is finally reported in the Fig.12 (bottom) as a function of the integration time: at each time, it is the ratio of the grey area of the middle panel (the cumulative of the positive lobe of the convoluted waveform) to the orange area of the top panel (the integral of the entire raw scintillation signal). As an example, aτexp =963 ns gives fint =92.77% for a 1000 ns integration time and fpurity =90.84%. The fint ·fpurity factor is then applied to the alpha pulse height spectra to properly asses the absolute PDE. The error on fint ·fpurity is computed by varying τpure and τexp and found to be <2%; as reported in Table 5their product is stable for τexp ranging from 900 to 1500 ns. Figure 13a shows the module geometrical acceptance as a function of the source position; the values are retrieved by the Montecarlo GEANT4 simulations described in Sect. 3.3. Two calibrated (by SPE charge) alpha spectra are also shown in Fig.13b; the difference of the detected photoelectron peak 123