First Demonstration of the Use of LG-SiPMs for Optical Readout of a TPC
A. Gola, K. Majumdar, G. Casse, K. Mavrokoridis, S. Merzi, L. Parsons Franca
PPrepared for submission to JINST
First Demonstration of the Use of LG-SiPMs for OpticalReadout of a TPC
A. Gola, a , b K. Majumdar, , c G. Casse, a , c K. Mavrokoridis, c S. Merzi, a , b , d and L. ParsonsFranca c a Fondazione Bruno Kessler (FBK), Via Sommarive, 18, 38123 Povo TN, IT b Trento Institute for Fundamental Physics and Applications (TIFPA),Via Sommarive, 14, 38123 Povo TN, IT c University of Liverpool, Dept. of Physics, Oliver Lodge Bld, Oxford Street, Liverpool, L69 7ZE, UK d University of Trento, Dept. of Physics, Via Sommarive, 14, 38123 Povo TN, IT
E-mail: [email protected]
Abstract:This paper describes a new method for optical readout of Time Projection Chambers (TPCs),based on the Linearly Graded Silicon Photomultiplier (LG-SiPM). This is a single photon-sensitivedetector with excellent timing and 2D position resolution developed at Fondazione Bruno Kessler,Trento (FBK). The LG-SiPM produces time-varying voltage signals that are used to reconstruct the3D position and energy of ionisation tracks generated inside the TPC.The TPC used in this work contained room-temperature CF gas at a pressure of 100 mbar,with two THGEMs to produce secondary scintillation light. A collimated Am source (Q α =5.486 MeV) was used to produce the ionisation tracks. The successful reconstruction of thesetracks is demonstrated, and the consistency of the methodology characterised through varyingthe geometry of the tracks within the TPC. Energy reconstruction and deposition studies are alsodescribed, demonstrating the feasibility of the LG-SiPM as a potential option for optical TPCreadout.Keywords: Time projection Chambers (TPC), Noble liquid detectors, Micropattern gaseous de-tectors, Photon detectors for UV, visible and IR photons (solid-state). Corresponding author a r X i v : . [ phy s i c s . i n s - d e t ] S e p ontents One of the most successful and widely used types of particle detector over the past few decades hasbeen the Time Projection Chamber (TPC). This type of detector can be operated in a number ofdifferent ways - for example, using a gaseous or liquid scintillator medium, or a combination of both.One of the most commonly used media in such detectors is liquid argon (LAr). There are a numberof LAr-based experiments currently operating around the world - for example, the three detectorsof the Short Baseline Neutrino Program [1]: SBND (which has 112 tons of LAr as its active TPCvolume), MicroBooNE (89 tons) and ICARUS-T600 (470 tons), as well as the single- (411 tons)and dual-phase (300 tons) ProtoDUNE experiments [2, 3]. TPCs will only continue to increase insize and sophistication in the near future, with the DUNE project proposing the use of four 17,000ton LArTPCs [4–10] and a 20,000 ton LArTPC being operated at the DarkSide-20k experiment [11].In a typical TPC, ionisation of the active scintillator material occurs as a particle travels through,producing scintillation light and free electrons. A micropattern charge amplifier, such as a ThickGas Electron Multiplier (THGEM), is typically used to multiply the electrons to increase the signal-to-noise ratio, and the charge signal is then usually collected on a segmented anode plane. Theelectron multiplication also acts a secondary source of scintillation light [12].– 1 –ue to the single-photon sensitivity of many modern optical imaging devices, observing thesecondary scintillation light is possible even at very low particle energies and signal-to-noise ratios,and can therefore offer lower energy thresholds than are possible using charge readouts. Dependingon the design and nature of the optical readout device(s), detector construction and operation mayalso be simplified considerably, leading to reductions in costs through the life of an experiment.Secondary scintillation light has previously been directly imaged using a range of devices:Geiger-mode avalanche photodiodes [13], silicon photomultipliers [14] and standard CCDs [15].More recently, the use of EMCCDs and Timepix3-based cameras have been successfully demon-strated by the ARIADNE Project at the University of Liverpool [16–18]. This R&D program isspecifically dedicated to developing innovative optical readout systems for use in current and futuredual-phase LArTPCs.In this paper, we introduce a new photo-detector: the
Linearly Graded Silicon Photomulti-plier (LG-SiPM) , along with a LG-SiPM-based camera system that can be used for the opticalreadout of TPCs that employ electron multiplication devices. This is a relatively new type of SiPMthat offers position sensitivity along both the x and y axes. It is also expected to provide excellenttiming performance, and therefore high resolution along the z axis. The Liverpool 40l TPC was used for these studies, and is shown in Figure 1. The 19 cm tall and17 cm diameter cylindrical field cage consists of 35 field-shaping rings, with a cathode grid locatedbelow them and an extraction grid positioned above. The TPC volume was filled with pure CF gasat a pressure of 100mbar, and the cathode and extraction grid were biased to -3.0 kV and -1.1 kVrespectively, giving an electric field of 95 V/cm within the TPC’s drift volume. A single 8 inchdiameter Hamamatsu R5912-20 PMT is positioned below the cathode grid, and is used to detectthe primary scintillation light for event timing and triggering purposes. It also acts as an additionaldetector for the secondary scintillation light produced in the THGEMs.A collimated Am α source (Q α = 5.486 MeV) is mounted on a rotating arm located 10 cmbelow the extraction grid. A rotation angle of 0° corresponds to the arm pointing at a tangent to thefield rings, and a maximum angle of 85° is possible. The source itself emits particles at an angle of45° downwards.Two identical THGEMs (Thick Gaseous Electron Multipliers) are located 1 cm above theextraction grid, with a separation of 1 cm between them. Each THGEM consists of a 1 mm thicksheet of FR4 with copper layers on the top and bottom planes. Approximately 23,000 evenly spaced500 µ m diameter holes are drilled through all three layers, covering an area of 150 cm . Eachhole has an additional 50 µ m etched rim which serves to increase the THGEM’s overall breakdownvoltage. The copper planes of each THGEM can be held at different voltages, as each of them isbiased from its own independent power supply. In these studies, the bottom THGEM’s bottom andtop planes have been held at -0.85 kV and 0 V respectively, with the top THGEM at 0 V (bottomplane) and +0.85 kV (top plane). – 2 – igure 1 . A 3D model of the Liverpool 40l TPC, with key components labelled. For clarity, the top THGEMand the outer wall of the vacuum chamber have been omitted. The entire internal TPC assembly described above sits inside the 40l ultra-high-vacuum cham-ber, which is capped by a 33.6 cm diameter conflat (CF) flange. Attached to and through this flangeare the various electrical feedthroughs for the field cage and THGEM voltages, the fluid in/outletfor filling and evacuating the chamber, the rotation arm for the α source, as well as a 10 cm diameterpipe, at the top of which is a borosilicate viewport. The viewport is 1 m from the top THGEMplane. – 3 –bove this viewport is the hardware for optical readout of the TPC: a Canon 85 mm f/1.4 cameralens coupled to the LG-SiPM mounting using a standard C-mount threaded ring. The mountingitself has been designed at the University of Liverpool and 3D-printed from ABS+, allowing theLG-SiPM to be held securely and stably as close as possible to the lens. Figure 2 shows the operating principle of the Liverpool 40l TPC. The key steps are discussed in themain text below.
Figure 2 . Schematic view of the operation principle of the Liverpool 40l TPC. Details are given in the maintext. Some detector components have been omitted from this 3D model for clarity.
When a particle enters the CF volume, it immediately produces prompt scintillation light (S1),which is detected effectively instantaneously by the PMT. This is therefore the known start time ofthe event. The particle also creates a track of ionised ions and free electrons as it moves through thevolume. Due to the electric field within the TPC field cage, these electrons drift upwards towards– 4 –he extraction grid and THGEMs. The high potential difference across each THGEM causes theelectrons to be accelerated through the THGEM holes, causing avalanches (electron multiplication)and the production of secondary light (S2). The amount of light produced is proportional to boththe number of THGEMs present and the potential difference between the top and bottom THGEMplanes. However, if this difference is too great, there may be enough ionisation within a hole to forma conductive plasma, leading to a discharge between the planes. The S2 light is emitted isotropically,and is therefore seen both by the PMT at the bottom of the detector and the the LG-SiPM at the topof the viewport above the vacuum chamber. The Linearly Graded Silicon Photomultiplier (LG-SiPM) has been designed as a new type ofposition-sensitive SiPM [19]. When a photon hits the sensor’s active area, the current generated bythe SiPM microcells is split into four outputs, from which it is possible to calculate the photon’s x and y coordinates, down to a theoretical spatial resolution equal to the size of the microcells - onthe order of 30 µ m. The LG-SiPM has a fast time response - typically on the order of a few tens ofns, and more recent SiPM designs show the possibility of reducing this response time down to lessthan 5 ns [20].The typical application of the LG-SiPM is for scintillation light readout in ultra-high-resolutionPositron Emission Tomography (PET). Millimeter [21] and sub-millimeter [22] spatial resolutionshave been recently demonstrated using 7 . × .
75 mm and 16 ×
16 mm devices respectively. Thesestudies also show the capability of the LG-SiPM in determining the precise position of a spatiallylocalised, short duration (on the order of tens of ns) source of light.The LG-SiPM’s characteristics make it a promising photo-detector in a variety of other ap-plications. The device is not a pixel detector, and so the signal produced by the four outputs arecontinuous in time. Therefore, if the active area is illuminated with a light source that changes posi-tion over time, the output signals of the LG-SiPM are expected to continuously change accordingly,and the equations of motion of the light source as it moves across the LG-SiPM’s active area canbe determined. This is of great interest in applications requiring light tracking with high sensitivityand high speed.At the same time, the Geiger mode operation of the LG-SiPM offers a high internal gain - onthe order of a few million units of charge per primary photoelectron. This allows it to reliably detectsingle photons without the need for an additional multiplication stage.
A schematic of the LG-SiPM microcell circuitry is shown in Figure 3. Each microcell contains twoquenching resistors, R Hq and R Vq , connected to horizontal and vertical buses respectively (depictedas brown and light blue lines on the schematic). Each bus is terminated with a resistive currentdivider composed of two resistors, the values of which are dependent on the bus location andorientation. For example, the resistors R l and R n − l − terminate the vertical bus at column l (out of n columns in total), and the horizontal bus at row t out of m total rows is terminated by resistors R t and R m − t − . The other ends of the resistors are connected to the two readout pads for each axis:– 5 –enoted as L and R for the horizontal axis, and T and B for the vertical. Each microcell requires abias voltage, V bias , which is provided to the entire structure from an external source. Figure 3 . Schematic of the microcell-level circuitry of the LG-SiPM. A full description is given in the maintext.
In the case of the vertical buses, the current generated by each microcell in response to the detectionof a single photon is divided between the two readout pads L and R according to the ratio ofthe current divider resistors. The difference between the conductances of R l and R n − l − changesproportionally to the column index l , and so the difference between the signals measured at the L and R pads is proportional to the x coordinate of the microcell with respect to the LG-SiPM activearea.The horizontal buses work equivalently to the vertical ones - bringing the signal to the T and B readout pads, which then provide information about the y coordinate of the fired microcell. Foreach vertical and horizontal bus, the parallel of the two current divider resistors is kept constantwith respect to the column or row index, so that - to a first approximation - each column or rowof microcells sees the same impedance towards the readout pads. This allows the horizontal andvertical buses to operate independently, with each one delivering only half of the charge generatedby the triggered microcell. In this way, pincushion distortions, which are typical of some othertypes of position-sensitive devices, are avoided. Two capacitors are also added in parallel to thecurrent divider resistors to improve the dynamic response of the detector.– 6 – .5 LG-SiPM Application for Optical TPC Readout Consider a simplified case of a single photon incident on a single microcell. The x and y coordinatesand the total charge Q of the microcell are related to the T , B , L and R readout signals by thefollowing: Q = L + R = T + B (2.1) x = R − LR + L (2.2) y = T − BT + B (2.3)In a more realistic scenario of a non-point-like light source incident on multiple microcells, equa-tions 2.2 and 2.3 now respectively describe the x and y coordinates of the centroid of the distribution,and Q is now also proportional to the number of incident photons. As the signals generated by eachmicrocell are combined at the four readout pads, it is no longer possible to obtain information aboutthe individual photons that make up the light distribution.For a moving light source, the equations of motion - x ( t ) and y ( t ) - are simply given by thevariation of equations 2.2 and 2.3 in time. This has been verified experimentally, using a simplifiedsetup comprising a light source collimated to a 1 mm diameter circular spot, and a motorisedtranslation stage, as shown in Figure 4. By moving the light spot from the left side of the LG-SiPMactive area to the right at a constant speed, the progression of the spot’s calculated x coordinate(shown by the waveform in Figure 4) is found.This setup involves a relatively slow rate of motion of the light source - limited by the translationstage’s motor - compared to the passage of particles through an actual particle detector. However,as noted previously, the LG-SiPM response is on the order of tens of ns, and it can therefore trackmuch faster movements than this setup allows for.The discussion above is directly analogous to the application of the LG-SiPM to optical TPCreadout: from the point of view of the LG-SiPM, the secondary scintillation light produced in theTHGEM is equivalent to a time-varying light source. This is depicted in Figure 5 for a simplified1D TPC containing an ionisation track.In the case of a generalised track geometry as shown, the first light is incident on the left sideof the LG-SiPM (negative x coordinate). The LG-SiPM then proceeds to generate a continuousanalog signal as the subsequent light reaches it. The specific shape and duration of the waveformat different points in time relate to the rate and intensity of the incident light, which in turn are areflection of the initial track.A simple extension of this to a 2D LG-SiPM - which provides waveforms for both x ( t ) and y ( t ) - shows that full 3D reconstruction of an ionisation track within the TPC is possible, with the– 7 – igure 4 . Schematic of the setup used to verify the response of the LG-SiPM to a moving light source. Theprogression of the light spot’s x coordinate, given by Equation 2.2 at any single point in time, is shown bythe waveform. The centre of the LG-SiPM is taken as the origin of the coordinate system. Figure 5 . The response of the LG-SiPM to an angled ionisation track in a simplified 1D TPC. The driftelectrons produce secondary scintillation light in the THGEM (not shown), and this light is then incident onthe LG-SiPM. The relative time and position of the incident light results in the specific duration and shapeof the output waveform. common t of the waveforms being directly related to z via the drift velocity of the material presentin the TPC volume.The results presented in this paper involved the use of a 2 × igure 6 . The 2 × The spatial resolution of the track reconstruction, which is an important measure of the effectivenessof the LG-SiPM readout, is dependent on how “clean” the overall signal is at the point of readout.One potential factor in this is the dark count rate (DCR) of the LG-SiPM. A dark count occurs whena microcell is fired by a carrier produced by thermal generation in silicon. It can be assumed thatthe DCR is a uniform distribution across the LG-SiPM’s active area, and so the light signal from atrack will always be superimposed on top of this flat background.The DCR is therefore directly related to the operating temperature of the LG-SiPM, whichis primarily affected by the ambient temperature. The bias voltage, V bias , that is provided to thedevice also has an impact - with a higher V bias leading to higher temperatures in the electronics,and therefore more thermal carrier generation. However, the signal from each microcell also scaleswith V bias , and so its value must be carefully chosen to maintain a balance between increased noiseand increased signal.For the purposes of the work presented in this paper, the LG-SiPM was operated at roomtemperature with V bias ranging from -34 to -36 V (noted where appropriate). However, studieshave shown that operating it at cryogenic temperatures can improve the performance by drasticallyreducing its DCR [23]. Data acquisition and readout of the 40l TPC is performed using a bespoke software frameworkdeveloped at the University of Liverpool. The PMT (1 channel) and LG-SiPM (6 channels) areall read out via a single 8-channel 12-bit 250 Ms/s CAEN V1720 digitiser, with the PMT channeladditionally being branched to a threshold discriminator. A global event trigger is issued from thisdiscriminator if the PMT (S1) signal exceeds a pre-defined threshold, and the trigger instructs thedigitiser to begin reading out all 7 channels synchronously. Each event window is 16 µ s long, giving4000 samples per channel per event. Data is stored in ASCII format, with one file per channel.– 9 – Results and Analysis
As noted in Section 2.5, the LG-SiPM array has 6 output channels: 3 along the x axis (denotedby Left [ L ], Central Horizontal [ C H ] and Right [ R ]) and 3 along y (Top [ T ], Central Vertical [ C V ]and Bottom [ B ]). A typical set of these output channels for a single α -induced event is shown inFigure 7. Figure 7 . The α -induced LG-SiPM signals from the x axis channels L , C H and R (left) and y axis channels T , C V and B (right). This event was recorded at a V bias of -34 V. The origin of the time axis corresponds tothe global event trigger as discussed in Section 2.6. Figure 8 shows the PMT and total LG-SiPM signals for the same α event (where “total” refers tothe summation over all 6 channels). It can be seen that the PMT detected the S1 signal at time“zero”, issuing the global event trigger and beginning the 16 µ s recording window for this event.The shapes of the PMT S2 and total LG-SiPM signals are very similar, as expected. Figure 8 . The PMT signal (black) from the same α event as displayed in Figure 7. Also shown (red) is thesummed LG-SiPM signal across all 6 channels, which has the same overall shape as the PMT signal - asexpected, since they both depict the same S2 scintillation light. – 10 – .2 Single Event 2D Track Reconstruction Following from the discussion in Section 2.5, the x − y projection of a single particle track can befound by considering the relative, and time-varying, ratios of the readout channels. in the case of 3readout channels per axis rather than 2, equations 2.1 to 2.3 become: Q = L + C H + R = T + C V + B (3.1) x = R − LR + C H + L (3.2) y = T − BT + C V + B (3.3)with each letter denoting the value of the signal on each corresponding channel at any given momentin time.Due to the intrinsic variability of the digitiser as well as photon dispersion within the detector,it is impractical to calculate the x and y coordinates for every single datapoint covered by the S2signal. However, using time “slices” representing many points at once helps to reduce the point-to-point variation considerably. The x − y reconstruction is therefore determined as a series of points,with each one representing a 80 ns (20 datapoint) wide time slice across all 6 channels, and its x and y coordinates calculated using equations 3.2 and 3.3. The values of L , C H , R , T , C V and B arethe respective mean values within the time slice.Figure 9 (left) shows the results of this for the α event previously displayed in Figure 7.Using the time ordering of the points, the general direction of the x − y projection is determinedto be from low to high y and high to low x . This is consistent with the known orientation of the α source relative to the LG-SiPM field of view. However, it can be seen that there are a number ofpoints at the end of the projection that do not follow the linear trend, but instead appear to pointback towards the centre of the LG-SiPM. This is a feature common to all events, and is due to thenon-zero “recharge time” of the LG-SiPM microcells.Each microcell require approximately 150 ns to discharge and reset after charging up due toan incident photon, and so there will be some charged microcells continuing to send signals for aperiod of time after the last photons from an event are incident on the LG-SiPM. This means that thelast few time slices used in the x − y reconstruction (the same ones that form the anomalous sectionin Figure 9 (left)) are in fact non-physical and should not be considered in the reconstruction.Figure 9 (right) shows the x − y reconstruction when the last 600 ns of the LG-SiPM signalis not used. It can be seen that the points now form a more linear distribution with a clear singledirection. Under the general operating principle of TPCs, the time between the S1 pulse and any point on theS2 pulse corresponds to the time taken for the free electrons at that corresponding point along the– 11 – igure 9 . (Left) The uncut x − y reconstruction for the α event displayed in Figure 7 using the methoddescribed in the main text. (Right) The reconstruction after performing a cut on the LG-SiPM rechargecomponent. In both cases, the line indicates a linear fit to all points using the Least Squares Method. track to drift up to the THGEM. This is related to the depth ( z coordinate) of that point on the trackthrough the the drift velocity, v d , of the scintillator medium at the particular drift field gradient andfluid pressure being used. For 100 mbar CF at a gradient of 95 V/cm, this velocity is on the orderof tens of mm/ µ s [24].From Figure 8 and the discussion in Section 2.6, the S1 pulse occurs at time “zero” in eachevent, and so the z coordinate at time t is simply given by: z t = v d × t (3.4)As with the x − y reconstruction, calculating the z coordinate of every datapoint in the S2 pulsewill result in an impractically noisy track due to the underlying point-to-point variability of thedata. Instead, the same time-slicing method is employed, using the same slice width as above andrepresenting each slice using its central time value.The x − z and y − z projections of the resulting 3D track reconstruction are shown in Figure 10.As with the x − y reconstruction, a cut on the last 600 ns of the S2 pulse (discussed previously inSection 3.2) has been applied. Although there is still some noticeable point-to-point variation, theoverall downward angle of the track is consistent with the known orientation of the α source. Theabsolute z values are also correct for the known position of the source relative to the top of the TPC(which is taken as z = The overlaid x − y projections of all reconstructed tracks in a single run is shown in Figure 11 (left).A cone of tracks can clearly be seen, with an opening angle and forward direction consistent with– 12 – igure 10 . The x − z (left) and y − z projections of the 3D track reconstruction, after a cut has been madebased on the LG-SiPM recharge time. In both cases, the line in a linear fit to all points using the LeastSquares Method. the shape and orientation of the α source and its collimator. Figure 11 . (Left) The overlaid x − y projections of all reconstructed tracks in a single run, taken at a V bias of -36 V and a 60° α source angle. (Right) The same overlay for only the filtered events - i.e. after applyingcuts on the PMT S2 signal area and reconstructed 3D track length. Figure 12 (black) shows the distribution of 3D track lengths across all reconstructed tracks. Thevast majority of tracks have lengths between 150 and 200 mm - as expected from a mono-energetic,single-species particle source embedded in a homogeneous scintillator medium.It can be seen that there are a small number of tracks with much shorter lengths than the– 13 – igure 12 . (Black) The (normalised) distribution of reconstructed 3D track lengths across all events in thesame run depicted in Figure 11. (Red) The same distribution for only the filtered events - i.e. after applyingcuts on the total PMT S2 signal area and reconstructed 3D track length. majority. It was found that these events also have significantly smaller S2 pulses than expected (interms of both duration and total area), suggesting that they are of much lower energy than the restof the events. The small S2 pulses make it difficult to accurately reconstruct the tracks - as can beseen on Figure 11 (left), where many of these short-length tracks appear to go backwards into thecollimator.It is presumed that such events are α particles that have become “caught” in the collimator afterbeing emitted, subsequently losing a significant fraction of their energy due to scattering and/orother interactions. Such events are not representative of the LG-SiPM characterisation being studiedhere, and can be removed by applying cuts on the PMT S2 signal area and reconstructed 3D tracklength. The effect of these cuts is shown in Figure 11 (right) and Figure 12 (red). For convenience,the events remaining after these cuts are referred to hereafter as the filtered events.The uncertainty in the x − y track reconstruction of a single event has been quantified as theRMS of the distance between each calculated point in the x − y plane and the linear fit to all points.Figure 13 shows the distribution of the RMS across the filtered events. For this particular run, themean error on the x − y track reconstruction is ≈ ( x , y , z ) and a 3D linear fit - is possible, in order to quantify the uncertainty in the overall 3Dtrack reconstruction. Work on this is still ongoing.The directionality of the reconstructed tracks can be emphasised by considering the track den-sity, shown in Figure 14. The data displayed were taken at two different orientations of the α source: 60° (left) and 45° (right). The difference between the two distributions is evidently clear,demonstrating the good position sensitivity and resolution of the LG-SiPM readout.– 14 – igure 13 . The (normalised) distribution of the x − y reconstruction error across the filtered events. Thiserror is calculated (per event) as the RMS of the distance between each calculated ( x , y ) point and the linearfit to all points. Figure 14 . The track density distribution for an α source angle of 60° (left) and 45° (right) across datasetsof filtered events. White colour indicates the highest density, and blue indicates the lowest. These data werecollected at a V bias of -35 V. As noted in Section 2.5, at a given V bias the total charge Q produced by the LG-SiPM (given in ADUby equation 3.1) is directly proportional to the number of incident photons. This in turn is related tothe number of electrons produced through the incident particle’s ionisation of the scintillator. Forthe single-species source used in these studies, the amount of ionisation - and therefore Q - is onlydependent on the particle energy.In principle, Q should take a single value, since the Am α particles are mono-energetic at– 15 –.486 MeV. However, in practice, detector effects and non-zero calorimetry resolution will broadenthe Q values into a distribution. Nonetheless, once this distribution of Q is found, a simple ADU-to-energy calibration can be used to scale the most probable Q value to the expected single α energy.The distribution of this reconstructed energy across many events is shown in Figure 15. Figure 15 . The distribution of reconstructed energy (red) across filtered events in a run taken at V bias = −
34 V.This distribution is that of the total LG-SiPM charge Q , given by equation 3.1, with the peak scaled to theenergy of Am α particles. For comparison, the black line indicates the distribution using all events in therun, showing the low energy events that are presumed to be α particles that were “caught” in the collimator. Based on the distribution’s Gaussian shape with a FWHM of ≈ ≈ purity (which could increasethe event-to-event variability of the number of electrons produced during ionisation), THGEM in-stabilities (which can affect the number of S2 photons produced), and the previously noted effectof the LG-SiPM’s DCR (which could increase the Q and contribute to the tail at higher energies).Further analysis to characterise and separate these effects is ongoing.The ratio of an event’s reconstructed energy to its previously reconstructed 3D track length givesthe energy deposition rate: dEdX . This quantity, which is also known as the “Linear Energy Transfer”(LET), is important for particle identification, as it is primarily a function of only the particle mass(and therefore, species) and momentum (which, in many particle detectors, is known or can berelatively easily determined).The distribution of the calculated LET is shown in Figure 16. It is seen to follow the expectedLandau-shaped behaviour, with a most probable value at ≈ The current LG-SiPM has shown great potential for optical readout of TPCs - having been able tosuccessfully reconstruct the 3D tracks of individual particles to a reasonable accuracy, and demon-strating good position resolution and a signal-to-noise ratio even when operated in room temperature– 16 – igure 16 . The distribution of the Linear Energy Transfer (LET) across filtered events, with a Landaufunction fitted to the distribution (black). The events in this data are the same as those presented in Figure 15. conditions and at a minimal V bias . However, these studies have brought to light certain limitationswith the current designs, which provide possible routes for future developments.Due to the way that the information from individual microcells is combined (and therefore lost)at the readout pads, the LG-SiPM is unable to correctly reconstruct events where a large numberof microcells are fired simultaneously. The two scenarios in which this was found to be particularproblematic are horizontal tracks and pileup events (where two or more tracks are present in thesame event window). In the case of horizontal (and near-horizontal) tracks, all of the S2 photonsarrive at the LG-SiPM at the same time, and so this is seen as a short, sharp signal in each readoutaxis instead of one with a flat shape and long duration (as would be seen by the PMT). The relatedsituation of pileup and near-simultaneous events sees the individual tracks merged into a single lightdistribution, with no clear separation between them that would allow their individual reconstruction.This imposes an upper limit on the event rate that the LG-SiPM is able to cope with.As noted, both of these limitations arise from the design of the LG-SiPM itself - simply becauseit is not a pixel detector, but instead combines the information across many pixels (microcells). Onepossible avenue for improvement to address this could be to add independent readout channelsfor specific regions of the active area. For example, if each quadrant of the 2 × x and y readouts alongside the 6 combined channels, signalscould be spatially localised in order to provide a more detailed picture of the track geometry. (Itis understood, of course, that the ultimate limit of this idea is when each microcell has its ownindividual readout channel, at which point the LG-SiPM effectively becomes a pixel detector.)As mentioned in Section 2.5, the LG-SiPM’s DCR would be greatly reduced by operation atcryogenic temperatures. However, the current design’s electronics are not cryogenic-compatible.Partially in response to this, a new production cycle of LG-SiPMs using FBK’s NUV-HDcryogenic technology is already underway. These detectors - which are expected to have a DCR of– 17 –lmost 0 Hz based on initial internal testing - will be able to stably operate at cryogenic temperatures,thus benefiting from reductions in the background power consumption and thermal variances in themeasurements. For the purposes of optical TPC readout, this will be reflected in less noisy outputsignals and smaller uncertainties in position and energy reconstruction.An additional benefit of the NUV-HD technology is a shift in the wavelength response ofthe LG-SiPM. The current device has a peak detection efficiency at 500 nm, with the efficiencydropping very sharply at shorter wavelengths. However, the PMTs that are commonly used inTPC-based particle detectors (including the 40l TPC used here) generally have the highest spectralresponse in the low 400 nm range, creating a mismatch between the devices. PMTs are commonlyused for detection of the S1 light pulse, which is a much smaller signal than the S2, and so it wouldbe more beneficial to operate the PMTs at their peak detection efficiency. This would then leavethe LG-SiPM with a reduced detection response. However, the NUV-HD technology being used inthe newest iteration of the LG-SiPM has a peak detection efficiency at 420 nm, matching almostperfectly with that of the R5912-20 Hamamatsu PMT used in the Liverpool 40l detector, and manyother similar devices used elsewhere.Alongside these large-scale changes to the underlying LG-SiPM design and manufacturing, small-scale fine-tuning of the components - such as capacitor and resistor values - can be of some additionalbenefit. With these changes, it may be possible to reduce the impact of the recharge time and otherartifacts, thereby reducing uncertainties on the position reconstruction. Digital post-processing ofthe readout signals can also potentially be implemented, either inline with the readout itself oras a pre-analysis software step, and this would benefit the signal-to-noise ratio and point-to-pointvariability. Post-processing could also improve the LG-SiPM’s response to pileup, allowing indi-vidual tracks to be distinguished and increasing the maximum event rate that the device is capableof operating at.It is anticipated that further testing and characterisation of both the current and future LG-SiPM de-signs will take place on the ARIADNE TPC. This detector is a 1-ton dual-phase LArTPC, and so willprovide a larger and more representative test-bed for the ongoing development of LG-SiPM-basedcameras for optical TPC readout. – 18 – cknowledgments The ARIADNE research program is proudly supported by the European Research Council GrantNo. 677927 and the University of Liverpool.The authors would like to thank the members of the Mechanical Workshop at the University ofLiverpool’s Physics Department and the researchers and technicians at Fondazione Bruno Kesslerfor their technical expertise and contributions.
References [1] P. A. N. Machado, O. Palamara, D. W. Schmitz, The Short-Baseline Neutrino Program at Fermilab,arXiv:1903.04608 [hep-ex][2] B. Abi et al., The Single-Phase ProtoDUNE Technical Design Report, arXiv:1706.07081[physics.ins-det][3] C. Cuesta, Status of ProtoDUNE Dual Phase, arXiv:1910.10115 [physics.ins-det][4] The DUNE Collaboration, Long-Baseline Neutrino Facility (LBNF) and Deep Underground NeutrinoExperiment (DUNE) Conceptual Design Report Volume 1: The LBNF and DUNE Projects,arXiv:1601.05471 [physics.ins-det][5] The DUNE Collaboration, Long-Baseline Neutrino Facility (LBNF) and Deep Underground NeutrinoExperiment (DUNE) Conceptual Design Report Volume 2: The Physics Program for DUNE atLBNF, arXiv:1512.06148 [physics.ins-det][6] James Strait et al., Long-Baseline Neutrino Facility (LBNF) and Deep Underground NeutrinoExperiment (DUNE) Conceptual Design Report Volume 3: Long-Baseline Neutrino Facility forDUNE June 24, 2015, arXiv:1601.05823 [physics.ins-det][7] The DUNE Collaboration, Long-Baseline Neutrino Facility (LBNF) and Deep Underground NeutrinoExperiment (DUNE) Conceptual Design Report, Volume 4 The DUNE Detectors at LBNF,arXiv:1601.02984 [physics.ins-det][8] The DUNE Collaboration, The DUNE Far Detector Interim Design Report Volume 1: Physics,Technology and Strategies, arXiv:1807.10334 [physics.ins-det][9] The DUNE Collaboration, The DUNE Far Detector Interim Design Report, Volume 2: Single-PhaseModule, arXiv:1807.10327 [physics.ins-det][10] The DUNE Collaboration, The DUNE Far Detector Interim Design Report, Volume 3: Dual-PhaseModule, arXiv:1807.10340 [physics.ins-det][11] C. E. Aalseth et al., DarkSide-20k: A 20 tonne two-phase LAr TPC for direct dark matter detection atLNGS The European Physical Journal Plus, Volume 133 (2018), Article Number 131[12] A. Buzulutskov, Electroluminescence and Electron Avalanching in Two-Phase Detectors, MDPIInstruments, Volume 4, Issue 2 (2020), Article Number 16[13] A. Bondar et al., First demonstration of THGEM/GAPD-matrix optical readout in two-phaseCryogenic Avalanche Detector in Ar, Nucl. Instrum. Meth. A, Volume 732 (2013), Pages 213-216[14] P. K. Lightfoot et al., Optical readout tracking detector concept using secondary scintillation fromliquid argon generated by a thick gas electron multiplier, Journal of Instrumentation, Volume 4, Issue4 (2009), Page P04002 – 19 –
15] K. Mavrokoridis et al., Optical Readout of a Two Phase Liquid Argon TPC using CCD Camera andTHGEMs, Journal of Instrumentation, Volume 9, Issue 2 (2014), Page P02006[16] K. Mavrokoridis et al., First Demonstration of Imaging Cosmic Muons in a Two-Phase Liquid ArgonTPC using an EMCCD Camera and a THGEM, Journal of Instrumentation, Volume 10, Issue 10(2015), Page P10004[17] D. Hollywood et al., ARIADNE - A Novel Optical LArTPC: Technical Design Report and InitialCharacterisation using a Secondary Beam from the CERN PS and Cosmic Muons, Journal ofInstrumentation, Volume 15, Issue 3 (2020), Page P03003[18] A. Roberts et al., First demonstration of 3D optical readout of a TPC using a single photon sensitiveTimepix3 based camera, Journal of Instrumentation, Volume 14, Issue 6 (2019), Page P06001[19] A. Gola et al., A novel approach to Position-Sensitive Silicon Photomultipliers: First results, 2013IEEE Nuclear Science Symposium and Medical Imaging Conference (2013 NSS/MIC), Pages 1-4[20] F. Acerbi et al., High Efficiency, Ultra-High-Density Silicon Photomultipliers, IEEE Journal ofSelected Topics in Quantum Electronics, Volume 24, Issue 2 (2018), Pages 1-8[21] A. Berneking et al., A new PET detector concept for compact preclinical high-resolution hybridMR-PET, Nuclear Instruments and Methods in Physics Research Section A: Accelerators,Spectrometers, Detectors and Associated Equipment, Volume 888 (2018), Pages 44-52[22] J. Du et al., Performance of a high-resolution depth-encoding PET detector module usinglinearly-graded SiPM arrays, Physics in Medicine & Biology, Volume 63, Issue 3 (2018), Page 035035[23] F. Acerbi et al., Cryogenic Characterization of FBK HD Near-UV Sensitive SiPMs, IEEETransactions on Electron Devices, Volume 64, Issue 2 (2017), Pages 521-526[24] S. R. Hunter, J. G. Carter, L. G. Christophorou, Electron motion in the gases CF , C F , C , andn-C F , Physical Review A, Volume 38, Issue 1 (1988), Page 58, Physical Review A, Volume 38, Issue 1 (1988), Page 58