Test-beam studies of a small-scale TORCH time-of-flight demonstrator
S. Bhasin, T. Blake, N. Brook, T. Conneely, D. Cussans, R. Forty, C. Frei, E. P. M. Gabriel, R. Gao, T. Gershon, T. Gys, T. Hadavizadeh, T. H. Hancock, N. Harnew, M. Kreps, J. Milnes, D. Piedigrossi, J. Rademacker, M. van Dijk
TTest-beam studies of a small-scale TORCHtime-of-flight demonstrator
S. Bhasin b,e , T. Blake c , N. Brook e , T. Conneely h , D. Cussans b , R. Forty f ,C. Frei f , E. P. M. Gabriel d , R. Gao a , T. Gershon c , T. Gys f , T. Hadavizadeh a ,T. H. Hancock a, ∗ , N. Harnew a , M. Kreps c , J. Milnes h , D. Piedigrossi f ,J. Rademacker b , M. van Dijk g a Denys Wilkinson Laboratory, University of Oxford, Keble Road, Oxford OX1 3RH, UnitedKingdom b H.H. Wills Physics Laboratory, University of Bristol, Tyndall Avenue, Bristol BS8 1TL,United Kingdom c Department of Physics, University of Warwick, Coventry, CV4 7AL, United Kingdom d School of Physics and Astronomy, University of Edinburgh, James Clerk MaxwellBuilding, Edinburgh EH9 3FD, United Kingdom e University of Bath, Claverton Down, Bath BA2 7AY, United Kingdom f CERN, EP Department, CH-1211 Geneva 23, Switzerland g CERN, EN Department, CH-1211 Geneva 23, Switzerland h Photek Ltd., 26 Castleham Road, St Leonards on Sea, East Sussex, TN389 NS, UnitedKingdom
Abstract
TORCH is a time-of-flight detector designed to perform particle identifica-tion over the momentum range 2 −
10 GeV / c for a 10 m flight path. The detec-tor exploits prompt Cherenkov light produced by charged particles traversinga quartz plate of 10 mm thickness. Photons are then trapped by total inter-nal reflection and directed onto a detector plane instrumented with customisedposition-sensitive Micro-Channel Plate Photo-Multiplier Tube (MCP-PMT) de-tectors. A single-photon timing resolution of 70 ps is targeted to achieve the de-sired separation of pions and kaons, with an expectation of around 30 detectedphotons per track. Studies of the performance of a small-scale TORCH demon-strator with a radiator of dimensions 120 × ×
10 mm have been performed intwo test-beam campaigns during November 2017 and June 2018. Single-photontime resolutions ranging from 104 . . . . ∗ Corresponding author
Email address: [email protected] (T. H. Hancock)
Preprint submitted to Elsevier February 19, 2020 a r X i v : . [ phy s i c s . i n s - d e t ] F e b een achieved for MCP-PMTs with granularity 4 ×
64 and 8 ×
64 pixels, respec-tively. Photon yields are measured to be within ∼
10% and ∼
30% of simulation,respectively. Finally, the outlook for future work with planned improvements ispresented.
1. Introduction
TORCH is a time-of-flight (ToF) detector designed to perform Particle IDen-tification (PID) at low momentum (2 −
10 GeV / c ) over a 10 m flight path [1, 2].The principle of operation is demonstrated in Fig. 1. TORCH exploits promptCherenkov photons produced by charged particles traversing a quartz plate of10 mm thickness, combining timing measurements with DIRC-style reconstruc-tion, a technique pioneered by the BaBar DIRC [3] and Belle II TOP [4, 5]collaborations. A fraction of the radiated photons are trapped by total in-ternal reflection, which then propagate to focusing optics at the periphery ofthe plate. Here a cylindrical mirrored surface maps the photon angle to aposition on a photo-sensitive detector; custom-designed Micro-Channel PlatePhoto-Multiplier Tube (MCP-PMT) detectors [6] are used to measure the timesof arrival and positions of each photon. Combined with external tracking in-formation, the spatial measurement allows the Cherenkov angle of the emittedphoton to be determined.The TORCH detector has been proposed for Upgrades Ib and II of the LHCbexperiment in order to improve the pion, kaon, and proton separation capabilityof the experiment in the 2 −
10 GeV / c range [7]. When installed in LHCb,TORCH will consist of eighteen identical 660 × ×
10 mm modules locatedroughly 9 . ∼
35 ps for a momentum of 10 GeV / c ,requiring a 10 −
15 ps time resolution for clean separation. This requires asingle-photon timing resolution of 70 ps, given around 30 detected photons pertrack.This paper builds upon the work presented in Ref. [2], in which a small-2 ollision Radiator (quartz plate)
PhotonDetector
Photon (one of many)
FocusingOpticsFocusing Optics(a) (b) P a r ti c l e PossiblePhoton Paths zy x y'
Figure 1: Schematics of a TORCH module demonstrating the principle of operation. (a)Total internal reflection traps Cherenkov light generated by a particle traversing the radiatorplate. (b) Upon reaching the focusing optics, the angle of the photon in the y − z plane ismapped to the y (cid:48) − coordinate on the detector, allowing θ C to be determined. Note that the y (cid:48) axis is rotated by 36 ◦ from the vertical ( y − axis). scale TORCH demonstrator was tested with a prototype MCP-PMT of circularconstruction, produced by Photek, UK. The same demonstrator has now beenverified with square 2-inch (nominal) Photek tubes, in two test-beam campaignsat the CERN PS (East Hall T9 facility) during November 2017 and June 2018.The TORCH demonstrator is described in Section 2. The test-beam infras-tructure is presented in Section 3. Section 4 discusses the set of data-drivencalibrations which were applied to the data. Results for the single-photon tim-ing resolution and photon counting efficiency are presented in Sections 5 and 6,respectively. Finally a summary and outlook for the future is given in Section 7.
2. The TORCH Demonstrator
The demonstrator consists of a 120 (width) ×
350 (height) ×
10 (thickness)mm radiator plate, optically coupled to a focusing block which has a cylindri-cally mirrored surface designed to focus 2 mm beyond the exit surface onto theMCP-PMT photocathode. The block has the same dimensions as it would havefor a full-sized module in LHCb, except having its width reduced to match the120 mm width of the plate. The radiator plate and focusing block assembly wasmounted into a rigid frame which allowed the angle of incidence of the beam3o be varied by tilting the demonstrator about the x -axis, seen in Fig. 1. Thecomplete structure was contained within a light-tight box and mounted upona translation table, allowing the module to be positioned in the x and y direc-tions with respect to the beam. Further details of the optical components andmounting mechanics can be found in Ref. [2]. In each of the two test-beam campaigns, the demonstrator was instrumentedwith a different two-inch square MCP-PMT with a 64 ×
64 anode pixelisation.The tubes were custom-designed for the TORCH project by Photek Ltd (UK)[6] and represent the final prototypes of a three-stage development process [8].Charge from the MCP electron avalanche is collected on a resistive layer (“sea”)inside the PMT vacuum, and capacitively coupled to the anode pads. Thisallows charge sharing to improve the spatial resolution beyond the anode-padpitch of 0.828 mm. In November 2017, the implemented MCP-PMT had a 4 × x, y (cid:48) ), where the coarse granularity was achieved by electricallygrouping pixels on an external Printed Circuit Board (PCB), connected to theanode pads using anisotropic conductive film. In June 2018 the granularity inthe x − direction doubled to 8 ×
64, which, with charge sharing, gives an effectivepixelisation which exceeds that required for optimal TORCH performance [9].For LHCb installation, a pixelisation of 8 ×
128 is planned.Both MCP-PMTs have an active area of 53 ×
53 mm , corresponding to ap-proximately half the width of the demonstrator. In both test-beam campaignsthe MCP-PMT was mounted between one side edge and the centre of the fo-cusing block, with the other half of the detector plane not being instrumented.In the 4 ×
64 MCP-PMT, the insulating layer which separates the resistive-seafrom the anode readout pads has a thickness of 0.5 mm. This results in a point-spread function at the pads of 1.80 ± ×
64 MCP-PMT has a 0.3 mm insulating layer, and results in a point-spread function atthe anode pads of 1.30 ± ×
64 MCP-PMT is around a factor twoless than for the 4 ×
64 device. Although the QE of the 8 ×
64 is not optimalfor reaching the desired number of photons per track, the performance of futuretubes is expected to improve with further iterations of development.
200 300 400 500 600 700 800
Wavelength (nm) . . . . . . . . Q u a n t u m E f fi c i e n c y ( % ) × × Figure 2: The quantum efficiencies of the 4 ×
64 (November 2017) and the 8 ×
64 (June 2018)MCP-PMTs, measured at CERN.
Readout electronics employing the NINO [10] and HPTDC [11] chipsets werecustom-developed for the TORCH project [12]. Due to the increased granularityof the 8 ×
64 MCP-PMT in the coarse-pixel direction, an entirely new readoutsystem was developed to replace that used for the 4 ×
64 device. Because ofdifferences in the size and shape of the boards, new holding mechanics werefabricated for the 8 ×
64 device, which introduced a 5 mm upwards offset of theMCP-PMT in the y (cid:48) − direction relative to the 4 ×
64 device.
As previosuly discussed, the Photek MCP-PMT was designed so that a singleincident photon will give hits on several neighbouring pixels [6]. This means thatthe 64 physical pixels in the y (cid:48) − direction can provide an effective granularity5f 128 pixels by exploiting charge sharing. In this way, to reconstruct singlephotons, hits are clustered according to the following criteria: • they must have the same x − coordinate (coarse pixel direction); • they must be adjacent neighbours in the y (cid:48) − coordinate; • the arrival of the hit must be timed within 1 ns of its neighbour.All three criteria must be met for any pair of hits to be included in the samecluster. However for the 8 ×
64 dataset, the criteria were slightly modified toaccount for a small fraction of dead channels: namely, if two clusters fall oneither side of a known dead channel and the hits neighbouring the dead channelfall within 2 ns of each other, then the clusters are merged. Cluster size isdetermined by the number of hits in the cluster, and the cluster position istaken to be the average position of the centroid of all the hits.
A simulation of the TORCH demonstrator has been developed, using opticalprocesses modeled by Geant4 [13]. Custom libraries were used to model thedetector response and readout, which take input from laboratory measurementsincluding the MCP-PMT quantum efficiency, gain, and point-spread function.Losses due to quartz surface scattering and Rayleigh scattering are modelled.The same simulation was used for both test-beam periods, but with differinginput from laboratory-measured parameters for the respective MCP-PMT used.
3. Test-beam Setup
In both test-beam campaigns, a 5 GeV / c beam was used, comprising approx-imately 70% pions and 30% protons. The TORCH demonstrator was positionedwith the beam striking half way down the radiator plate, 5 mm from the edge(below the MCP-PMT), and tilted back from the vertical by 5 ◦ . This geomet-rical configuration ensured that the Cherenkov pattern was well contained onthe MCP-PMT detector surface. 6he same beam-line infrastructure was installed for both campaigns, dis-played schematically in Fig. 3. A pair of identical timing stations, T1 andT2, spaced approximately 11 m apart, was used to provide a time reference forTORCH. Each station, oriented at 49 ◦ to the beam, consisted of a 100 mmlong, 8 × borosilicate bar in which Cherenkov light was generated fromtraversing particles. A single-channel MCP-PMT detected the direct photonsand provided a precise timing signal. The signals were injected into the TORCHelectronics and read out simultaneously with the rest of the data. By combiningsignals from both stations, a time of flight measurement could be made inde-pendently of TORCH, providing a cross-check of PID for the particle traversingthe TORCH prototype. Additionally, each station had a pair of scintillatorsproviding an 8 × coincidence. Requiring a signal in both scintillatorsnarrowed the beam definition accepted by the trigger and improved the resolu-tion of the time reference. The timing power of the stations is demonstrated inFig. 4, which shows clearly the separation of pions and protons in the beam. Inaddition, a pair of threshold Cherenkov counters filled with CO at 2 . ~11m T2TORCHTelescopeT1C2C1Beam Figure 3: A schematic showing the beam-line configuration. C1 and C2 are Cherenkov coun-ters. T1 and T2 are timing stations spaced approximately 11 m apart.
An EUDET/AIDA pixel beam telescope [14] was also installed in the beam-line, consisting of six 18 . µ m pitch sensors (Mimosa26). The telescope allowedan accurate measurement of the beam profile incident on TORCH, even thoughan event-by-event synchronization was not possible. Fig. 5 shows the beamprofile measured by the telescope when extrapolated to the TORCH radiator,giving an RMS spot size of 2 . ± .
02 mm in x and 2 . ± .
02 mm in y . Thebeam divergence was measured to be 5 . ± . . ± . x and7 Time of Flight (ns) E v e n t s ( × ³ ) Figure 4: The time-of-flight difference (T2 − T1) over an 11 m flight path, showing the pioncontribution in red (dotted) and the proton contribution in blue (dashed). Note that the zerotime is arbitrary and, as defined, pions arrive later in time than protons. The horizontal barsdenote the bin width. y , respectively.Triggering of the TORCH readout and telescope was provided by an AIDA-2020 Trigger Logic Unit (TLU) [15]. The new beam-line infrastructure alloweda large increase in achievable data rate with respect to [2]. By providing theindependent source of PID, the Cherenkov counters allowed T1 to be removedfrom the trigger, with T2 alone being used as a time reference. This led to awider beam profile that could be triggered upon, significantly increasing the ac-ceptance and trigger rate. Comparing the PID information from the Cherenkovcounters with the PID from ToF, the purities of the pion and proton samplesfrom the Cherenkov counters were approximately 94% and 82% in November2017, and 98% and 96% in June 2018, respectively.
4. Calibrations
Two data-driven calibrations were applied to the data to correct the timingof the MCP-PMT output signals, the first to account for timewalk in the NINOchip, and the second to correct for integral non-linearity in the HPTDC.The first correction accounts for timewalk of the NINO (i.e. differences in8 a r ti c l e s − x Coordinate (mm) − − y C oo r d i n a t e ( mm ) Figure 5: The beam profile in x and y measured by the telescope when extrapolated to theTORCH radiator. The outer structure of the shape maps the beam profile and is attributedto scattering. The central distribution is the region defined by the scintillator trigger. timing due to variations of pulse amplitude), and is adapted from the data-driven method employed in Ref. [2]. The first stage in the calibration processis to define an MCP-PMT photon cluster. Assuming each cluster correspondsto a single photon, the hit pixels within that cluster should have simultaneousrecorded times, and any difference ∆ t i,j between pairs of channels i, j would bea consequence of time slew. The NINO utilizes a time-over-threshold technique,outputting a binary signal with a width defined by the rising and falling edgesof the MCP-PMT input pulse when passing an adjustable threshold. The signalwidth w i for a channel i is hence related to the input pulse amplitude. Thisintroduces a relationship between ∆ t i,j and the corresponding pulse widths,which can be parameterised as:∆ t i,j = t i − t j = F ( w i , w j ) , (1)where t i,j are the recorded times of the hits on channels i, j , and F is chosen to bea 2-dimensional function of quadratic form, with those coefficients determinedby a fit to pairs of hits from the same cluster. The constant term also correctsfor the relative delay between individual pixel timing offsets ( t ’s). In this way,the parameters of F can be determined for all pairs of channels during a data9un. Thereafter a correction is made to the measured arrival time of each singlepixel hit according to its measured pulse width. This method assumes the timewalk of each individual pixel is uncorrelated with all the others, and improvesthe method employed in Ref. [2] by comparing all pairs of hits, rather thanparameterising and correcting only next-to-nearest neighbours.The second calibration accounts for non-linearity in the HPTDC chip, wherethe bins used to digitise the data are not equally spaced in time [11], leading tointegral non-linearity. Several large dedicated calibration datasets were takento allow a code-density test [16] to be performed to correct for this effect.A calibration step which is presently missing is the so-called charge-to-widthcalibration, which would allow a more accurate measurement of the amountof charge collected in any given pixel hit as a function of the width of theoutput pulse. This in turn would allow more accurate cluster centroiding tobe performed. This calibration requires a dedicated laboratory-based charge-injection system, and this is currently under development.
5. Single-Photon Time Resolution
Figure 6 shows the uncorrected distributions (pixel maps) of hits on theMCP-PMTs from the two run periods for pions and protons combined, takenwith the beam positioned at the vertical mid-point of the radiator plate, 5 mmfrom the edge below the MCP-PMT. Bands can be seen, corresponding to differ-ent photon paths within the radiator plate. The empty bins in Fig. 6b indicatedead channels. These are attributed to broken wire bonds of the NINO elec-tronics board, an issue which has been resolved in subsequent iterations.The single-photon time resolution of the demonstrator can be measured bycomparing the time at which a photon is detected to that predicted from theTORCH reconstruction algorithm [2]. The algorithm determines the photonpath in the radiator plate and the Cherenkov angle from the position of thetrack entry point, the track direction, and the position of the photon hit onthe MCP-PMT. Combining this with knowledge of the primary particle speciesand its momentum, the time of propagation can then be calculated through the10 x Coordinate (pixels)(a)(b) x Coordinate (pixels) y ' C oo r d i n a t e ( p i x e l s ) H it C oun t H it C oun t y ' C oo r d i n a t e ( p i x e l s ) Figure 6: Uncorrected MCP-PMT hit-maps for combined pions and protons showing (a) the4 ×
64 and (b) the 8 ×
64 MCP-PMT (November 2017 and June 2018 test-beam periods,respectively). The black bins in (b) indicate a dead channel. Note that although both beamtests used an MCP-PMT with 64 pixels in the y (cid:48) coordinate, the top part of the detector wasnot illuminated and so the distributions have been truncated. intermediate steps of determining the phase and group refractive indices. Notethat the nominal values of beam position and incident angle are used in thereconstruction, leaving any finite beam width and divergence to be accountedfor statistically, as described below.For each column of pixels, the measured arrival time can be plotted againstthe y (cid:48) (finely-granulated) pixel number. Fig. 7a shows an example distributionfor the 8 ×
64 MCP-PMT, selected for protons only, with the predictions fromthe reconstruction algorithm overlayed. In calculating the predicted time in thereconstruction, each photon is treated individually and its energy calculated111]. The distinct bands seen in the figure correspond to the different ordersof reflection from the side faces of the demonstrator, illustrated schematicallyin Fig. 7b. This clearly demonstrates that photon paths in the radiator plateare well separated. Note that within a given order of reflection for a specificset of track parameters, the measured y (cid:48) pixel coordinate is correlated to theCherenkov photon energy, with the finite pixel size contributing to the chromaticuncertainty.For those photon hits which had either no reflections off a side edge or whichonly had a reflection off the edge below the MCP-PMT (corresponding to orders0 and 1 (cid:48) in Fig. 7b), a residual distribution (i.e. the measured minus predictedtimes of arrival) is constructed. The residual distributions of individual bins in y (cid:48) are first fitted to determine the resulting mean. These means are expectedto be offset with respect to each other due to chromatic dispersion, and thusare corrected by offsetting the photon arrival times within each bin by themean of the bin. Recombining the bins then gives the final fitted distribution.The sigmas are also dependent on photon energy, hence the measurements areaveraged. 12 ' Coordinate (pixels) H itti m e - T i m e R e f e r e n ce ( n s ) P ho t on C oun t RadiatorPlatePhoton Hit
MCP
Particle Hit ' 2 1' ' 1 0 (b)(a) Figure 7: (a) The distribution of detected photons as a function of y (cid:48) (finely-granulated)pixel number and arrival time, for column 5 of the 8 ×
64 MCP-PMT (see Fig. 6b). Here theselection is for protons. The three distinct peaks correspond to different orders of reflectionoff the sides of the radiator plate, and the overlaid lines show the predicted time of arrivalfor each order. The reduced population of hits for pixels beyond 32 is due to the differentthreshold settings for a pair of adjacent NINO chips which read out the MCP-PMT column.(b) A schematic of the photon paths assigned in the reconstruction which correspond to theoverlaid lines in (a), labelled according to the number of photon side reflections. Paths firstreflecting off the edge closest to the beam are shown by dotted lines, and their labels haveprimes. −600 −400 −200−800 0 200 400 600 800 1000 (ps)- T predicted T measured P ho t on C oun t Figure 8: An example residual distribution for orders of reflection 0 and 1 (cid:48) , as indicated inFig. 7b. The data are fitted with a Crystal Ball function with parameters µ = 0 ± σ = 115 ± α = − . ± .
03, and n = 4 . ± . The time spread of the residual distribution, σ Total , is a combination ofseveral factors which must be subtracted in quadrature to give the intrinsictiming resolution. This is given by σ = σ + σ + σ , (2)where σ TORCH is the intrinsic single-photon timing resolution which we wishto determine, σ Beam is the spread in the residual distribution resulting fromthe finite width and divergence of the incident beam, and σ TimeRef is the timeresolution of the T2 station that provides the reference time for TORCH. Thelatter two contributions are discussed below.
The beam-spread contribution : The contribution to the time spread due tothe beam profile, σ Beam , is determined from simulation. The simulation is run14ith the spread in beam position and divergence as measured by the telescope,and then for a beam with no spread. In each case, a residual distribution isconstructed the same way as for data, and the width of the two compared. Thevalue of σ Beam is found to be 14 ± The time reference contribution : The T2 station provides the time referencewith respect to which the Cherenkov photons in TORCH are measured. Thetime resolution of the stations is demonstrated in Fig. 4, showing the (T2 − T1)time difference in June 2018 data. The pion and proton beam contributionsare clearly separated. To determine the resolution of the downstream stationT2 independent of TORCH, it is assumed the behaviour of the pair of timereference stations T1 and T2 is identical before their signals propagate to theTORCH readout, however the timing of upstream station T1 is degraded due to ∼
11 m of cable length. When this contribution is isolated, the resolution of T2is determined to be σ TimeRef = 43 . ± . . ± . σ TORCH is determined separately foreach MCP-PMT column and for each incident particle species. Unfortunately asignificant pollution of pions is observed in the proton sample for the November2017 dataset due to the Cherenkov counters being non-optimally tuned, so onlyphotons resulting from an identified pion are used in this case. This results infour measurements for the 4 ×
64 MCP-PMT, presented in Table 1, and 16 forthe 8 ×
64, shown in Table 2.MCP Column σ TORCH (ps)1 112 . ± .
42 114 . ± .
43 104 . ± .
44 111 . ± . Table 1: The single-photon time resolutions for pions measured for the 4 ×
64 MCP-PMTin November 2017. The MCP column numbers match those shown in Fig. 6a. The quoteduncertainties are purely statistical. σ TORCH σ TORCH
Column Pions (ps) Protons (ps)1 110 . ± . . ± .
42 101 . ± . . ± .
43 101 . ± . . ± .
44 105 . ± . . ± .
45 83 . ± . . ± .
46 101 . ± . . ± .
27 90 . ± . . ± .
48 112 . ± . . ± . Table 2: The single-photon time resolutions for pions and protons measured for the 8 × The measurements from the two test-beam periods are generally similar,with resolutions σ TORCH between 100 and 110 ps typically observed. The overalltrend of enhanced performance in the 8 ×
64 MCP-PMT with respect to the 4 × x − direction due to the doublingof the number of pixel columns. Columns 5 and 7 stand out in particular forthe 8 ×
64 MCP-PMT data, with measured resolutions of order 90 ps. Thisresults from the application of better calibration corrections for these columns.It is noted that six of the eight columns for the 8 ×
64 MCP-PMT give betterresolutions for pions than protons, an effect which is attributed to a residual pionpollution in the proton sample in the June 2018 data. In this case a fraction ofpions will be falsely reconstructed as protons, resulting in an incorrect predictedtime.As indicated in Fig. 7a, orders 0 and 1 cannot be distinguished in data.The difference in time of arrival between the two orders from the simulationranges from 5 −
30 ps, varying with the x − position of the hit. This will widenthe residual distribution, and leads to a slightly degraded time resolution beingmeasured than with only a single order of reflection. However, as the effect isphoton-energy dependent, no attempt has been made to subtract this contribu-tion.Incorporating a charge-to-width calibration for the NINO in addition to the16ata-driven approach here employed would allow a charge-weighted average ofthe time and position of each cluster to be determined. This would improve theresolution further, and closer to the desired 70 ps.
6. Photon Counting
The photon counting efficiency of the demonstrator is determined by count-ing the number of detected clusters and comparing to the number expectedin simulation. Monte Carlo samples corresponding to 10000 incident pions at5 GeV/ c were used for each detector configuration. Figure 9 shows the distribu-tions of number of photons seen and expected in data and simulation, respec-tively, for the two MCP-PMT arrangements. The arithmetic mean numbers ofmeasured photons are compared in Table 3. A negligible difference is observedin counting efficiency between pions and protons, hence no selection is made inthe data based on the species.MCP-PMT Data Simulation Ratio4 ×
64 7 . ± .
03 8 . ± .
03 0 . ± . ×
64 3 . ± .
02 5 . ± .
03 0 . ± . Table 3: The arithmetic means of the photon-yield distributions shown in Fig. 9 and the ratioof data compared to simulation. The quoted uncertainties are purely statistical.
The reduced number of clusters observed for the 8 ×
64 MCP-PMT withrespect to the 4 ×
64 device in both data and simulation is expected, given thatthe 8 ×
64 device has a significantly lower quantum efficiency (as seen in Fig. 2).The photon yield in the 4 ×
64 MCP-PMT agrees within 10% of the simulation,however for the 8 ×
64 device, a ∼
30% loss in data is observed. The yieldsdepend strongly on the MCP-PMT gains and the NINO thresholds, the bestestimates of which were used in the simulation . The observed discrepancies areattributed to uncertainties due to small signals arising from charge sharing, for The gain of the 4 ×
64 has been measured to be 1 . × electrons, whilst the nominalvalue of the gain for the 8 ×
64 is 1 . × electrons, taken from the Photek data sheet. Thesevalues were used respectively, along with a NINO threshold value of 30 fC.
10 20
Number of Clusters F r e qu e n c y ( A . U . ) (a) Photon Count N o r m a li s e d F r e qu e n c y (b) Figure 9: The number of clusters seen in data (histograms) and simulation (points) for (a) the4 ×
64 and (b) the 8 ×
64 MCP-PMT. The smaller number of photons observed in simulationfor the 8 ×
64 is due to its lower quantum efficiency. which the systematics from the NINO threshold values are significant. Futurelaboratory work will therefore focus on improving the efficiency and calibrationof the MCP-PMTs and the electronics.
7. Summary and Future Plans
Studies of a small-scale TORCH demonstrator with customised MCP-PMTsand readout electronics have been performed during two test-beam periods inNovember 2017 and June 2018. Single-photon time resolutions ranging from1804 . . . . ×
64 and 8 ×
64, respectively. The improvement for the 8 × −
40% of the 70 ps targeted. The photon yields show a strong depen-dence on the MCP-PMT quantum efficiency, and also highlight future work thatis required to better understand the factors associated with the operational pa-rameters of the MCP-PMT, the properties of charge sharing, and the calibrationof the readout electronics.A half-sized LHCb demonstrator module with a 660 × ×
10 mm radiatorplate has been constructed and is currently being evaluated [18]. The demon-strator has been instrumented with the same 8 ×
64 MCP-PMT and readout elec-tronics as for the June 2018 beam test, alongside a second identically-configured8 ×
64 MCP-PMT with an improved quantum efficiency. This will allow tim-ing resolution studies and photon-yield measurements with improved photonstatistics. Analysis is underway and will be the subject of a future paper.
Acknowledgements
The support is acknowledged of the Science and Technology Research Coun-cil, UK, (grant number ST/P002692/1) and of the European Research Councilthrough an FP7 Advanced Grant (ERC-2011-AdG 299175-TORCH). The au-thors wish to express their gratitude to the CERN EP-DT-EF group, SimonPyatt of the University of Birmingham, and to Gale Lockwood of the Universityof Oxford/RAL for their efforts in wire bonding the NINO chips. We also thankAndre Rummler for his support on the test-beam telescope. We are grateful toNigel Hay, Dominic Kent and Chris Slatter of Photek, and to Jon Lapington ofthe University of Leicester for their work on the MCP-PMT development. TBacknowledges support from the Royal Society, UK.
References [1] M. Charles, R. Forty, TORCH: Time of flight identification with Cherenkovradiation, Nucl. Inst. & Meth. in Phys. Res. A 639 (2011) 173 – 176. doi: .[2] N. Brook, et al., Testbeam studies of a TORCH prototype detector, Nucl.Inst. & Meth. in Phys. Res. A 908 (2018) 256 – 268. doi:10.1016/j.nima.2018.07.023 .[3] I. Adam, et al., The DIRC Particle Identification System for the BABARExperiment, Nucl. Inst. & Meth. in Phys. Res. A 538 (2005) 281 – 357. doi:10.1016/j.nima.2004.08.129 .[4] T. Abe, et al., Belle II Technical Design Report (2010). arXiv:1011.0352 .[5] J. Fast, The Belle II imaging Time-of-Propagation (iTOP) detector, Nucl.Inst. & Meth. in Phys. Res. A 876 (2017) 145 – 148. doi:10.1016/j.nima.2017.02.045 .[6] T. Conneely, et al., The TORCH PMT: a close packing, multi-anode, longlife MCP-PMT for Cherenkov applications, Journal of Instrumentation 10(2015) C05003. doi:10.1088/1748-0221/10/05/C05003 .[7] The LHCb Collaboration, LHCb Experiment, Physics case for an LHCbUpgrade II, Tech. Rep. LHCb-PUB-2018-009. CERN-LHCb-PUB-2018-009, CERN, Geneva (May 2018).URL https://cds.cern.ch/record/2320509 [8] N. Harnew, et al., Status of the TORCH time-of-flight project, Nucl. Inst.& Meth. in Phys. Res. A (2018) doi:10.1016/j.nima.2018.12.007 .[9] L. Castillo Garc´ıa, et al., Development, characterization and beam tests of asmall-scale TORCH prototype module, Journal of Instrumentation 11 (05)(2016) C05022–C05022. doi:10.1088/1748-0221/11/05/c05022 .[10] F. Anghinolfi, et al., NINO: an ultra-fast and low-power front-end am-plifier/discriminator ASIC designed for the multigap resistive plate cham-ber, Nucl. Inst. & Meth. in Phys. Res. A 533 (2004) 183 – 187. doi:10.1016/j.nima.2004.07.024 . 2011] A. Akindinov, et al., Design aspects and prototype test of a very preciseTDC system implemented for the Multigap RPC of the ALICE-TOF, Nucl.Inst. & Meth. in Phys. Res. A 533 (2004) 178 – 182, proceedings of theSeventh International Workshop on Resistive Plate Chambers and RelatedDetectors. doi:10.1016/j.nima.2004.07.023 .[12] R. Gao, et al., Development of TORCH readout electronics for customisedMCPs, Journal of Instrumentation 11 (04) (2016) C04012–C04012. doi:10.1088/1748-0221/11/04/c04012 .[13] J. Allison, et al., Recent developments in Geant4, Nucl. Inst. & Meth. inPhys. Res. A 835 (2016) 186 – 225. doi:10.1016/j.nima.2016.06.125 .[14] I. Rubinskiy, An EUDET/AIDA Pixel Beam Telescope for Detector De-velopment, Physics Procedia 37 (2012) 923 – 931, proceedings of the 2ndInternational Conference on Technology and Instrumentation in ParticlePhysics (TIPP 2011). doi:10.1016/j.phpro.2012.02.434 .[15] P. Baesso, et al., The AIDA-2020 TLU: a Flexible Trigger Logic Unit forTest Beam Facilities, Journal of Instrumentation (2019) (in press) doi:10.1088/1748-0221/14/09/P09019 .[16] S. Liu, et al., LUT-based non-linearity compensation for BES III TOF’stime measurement, Nuclear Science and Techniques 21 (1) (2010) 49. doi:10.13538/j.1001-8042/nst.21.49-53 .[17] T. Skwarnicki, A study of the radiative cascade transitions between theUpsilon-prime and Upsilon resonances, Ph.D. thesis, Institute of NuclearPhysics, Krakow (1986).URL [18] T. H. Hancock, et al., Beam tests of a large-scale TORCH time-of-flightdemonstrator, To be published in Nucl. Inst. & Meth. in Phys. Res. A(2019) doi:10.1016/j.nima.2019.04.014doi:10.1016/j.nima.2019.04.014