LHCb 2000-32, CALO 9 June 2000 Results of a tagged photon test beam for the Scintillator Pad Detector. Llu s Garrido David Gascón Ramon Miquel Daniel Peralta Universitat de Barcelona, ECM department E-08028 Diagonal 647 (Barcelona) IFAE, Universitat Aut onoma de Barcelona E-08193 Bellaterra (Barcelona) Abstract The Scintillator Pad Detector belongs to the ECAL system and has to separate photons and electrons at the level-0 of the ECAL trigger. It has its own read-out system and it is positioned just before the preshower (PS). This note reports on the test beam for the SPD and shows its design, Monte Carlo simulation and results. The tagged photon beam allowed testing photon signals at the SPD. These signals are mainly due to pair production inside the scintillator and to charged particles generated in the shower at PS and ECAL (backsplash). The results show that both effects do not introduce large inefficiencies in photon-electron separation. 1 Introduction ECAL and PS cannot, by themselves, distinguish between high-pt photons and electrons at the level-0 of the LHCb trigger system. The SPD is contemplated within the Calorimeter group of the LHCb collaboration as the solution to perform this separation. The SPD will be just before the PS and will have the same acceptance and granularity. The SPD present design is basically the same of the scintillator sector of the PS, that is a matrix of scintillator pads with an optical fiber inside the plastic used to collect and guide the scintillation light. A detailed description for the calorimeter and trigger systems can be found in [1]. Charged particles will, and neutrals will not, produce ionization in the scintillators. If the scintillation light is collected and detected, then, only a logical 1
signal associated to the 0-MIP or 1-MIP case is needed from the SPD to complete the level-0 trigger system for the ECAL. Some processes can cause a photon to deposit indirectly energy in the scintillator and, then, the photon could be misidentified as an electron. From the physical point of view, these are the main sources of photon misidentification: ffl Pair production before SPD. This process depends on the amount of material before SPD in LHCb and will not be treated in this note. ffl Pair production, or any other process that produces charged particles, inside the SPD. In the note, it will be referred as Interaction. ffl Charged particles generated at the photon electromagnetic shower and moving backwards. In the note, it will be referred as Backsplash. The aim of the test beam is to collect signals from electrons and photons crossing a scintillator pad in an environment that includes the PS and ECAL and measure the distribution of deposited energies. For photons, interaction and backsplash contributions to the signals will be disentangled. The distributions obtained this way will be used to determine the SPD performance. The note is described here: ffl Section 2 describes the Monte Carlo prediction for the SPD photon and electron signals. ffl Section 3 details the test beam design. ffl Section 4 introduces the strategy used to analyze the data, and discusses some MC tests. ffl Section 5 includes the details of the results found. ffl Section 6 concludes the note. 2 Monte Carlo simulation Using the stand-alone detailed simulation of Calorimeters (SPD+PS+ECAL+- HCAL) in experimental area X7, SPD photon and electron signals have been simulated. In each event, a single photon or electron is fired at the center of an SPD cell and the energy deposited in that SPD cell is saved. No forward simulation about the scintillation process and light collection and detection is included. Figure 1 shows the distribution of the deposited energies in the SPD by 10 GeV photons and electrons. In the photon case, the distribution of deposited energies falls rapidly and is almost zero above 4 MeV. Most of the photons do 2
not deposit energy (therefore the large peak at zero) and the rest deposit energy through pair production inside the SPD or through backsplash particles. In the electron case, the peak of the distribution is very close to 1.65 MeV, there is a small tail to low energies down to 1.4 MeV, and a longer tail to higher energies. In this note, whenever we use MIP units for the deposited energy in SPD, 1 MIP signal is an amount of energy equivalent to the position of the peak in the electron signal distribution, that is, about 1.65 MeV. Applying a cut in the SPD deposited energy, the performance of the SPD in separating photons and electrons can be estimated. Figure 2 shows the fraction of photons and electrons that deposite more energy in the SPD than a certain cut. These fractions reflect the probability of misidentifying an electron as a photon or a photon as an electron depending on the SPD energy cut. The probability to misidentify a photon has been studied for different photon energies and the sources of misidentification, interaction with SPD and backsplash, have been split apart. As seen in figure 3, photon misidentification due to interaction has a soft dependence with photon energy for photon energies above 10 MeV. The probability of photon misidentification due to backsplash clearly increases with the photon energy and, for photons with energy below 1 GeV, is negligible compared to misidentification due to interaction. Table 1 shows numerically the results for three different photon energies and two different cuts. The simulation also shows that deposited energies by interaction and backsplash are not correlated (see figure 4) and are peaked at zero and fall sharply (fig. 1), hence, interaction and backsplash can be treated as small independent effects and their contributions simply added. fl energy SPD cut I mis B mis (GeV) (MIPs) (%) (%) 10 0.70 1.15 ± 0.08 0.28 ± 0.04 20 0.70 1.13 ± 0.07 0.33 ± 0.04 50 0.70 1.10 ± 0.11 0.55 ± 0.08 10 0.35 1.45 ± 0.09 0.84 ± 0.04 20 0.35 1.47 ± 0.08 0.96 ± 0.07 50 0.35 1.50 ± 0.13 1.30 ± 0.12 Table 1: Fraction of photons (%) of three different energies (10, 20 and 50 GeV) misidentified due to interaction (I mis ) and backsplash (B mis ) when a cut is applied in the SPD deposited energy. A MIP corresponds to 1.65 MeV of deposited energy. 3
3 Test Beam Design The tagged test beam allows working either with photons or electrons. Photons are produced by electrons crossing a lead radiator. The magnet deflects the electrons and the tagger selects the electrons by their energy. For the used 1.637T m magnet, the deflection can be estimated by eq. (1). (mrad) ß 500 p(gev ) To select electrons between 2 GeV and 6 GeV (using a pure electron beam of 20 GeV or 50 GeV) the tagger must be a 50 cm long scintillator bar positioned at 3 m beyond the magnet and at 25 cm from the beam line. The veto before the SPD rejects electrons created by pair production in the transit of the photons from the radiator to the veto. The distance between veto and SPD is then used to know the number of photons hitting the detector through the seen electron pairs in the SPD. The large distance between veto and SPD decreases the chances that a charged backsplash particle hits the veto. ECAL is included in the trigger system to know if any high energy particle (above 10 GeV) crossed its central cell. It is done in order to suppress events where the particle did not cross the SPD due to the electron beam angular divergence or the multiple scattering in the radiator. For photons, magnet on and radiator in, the trigger asked for coincidence of beam counter (1), tagger (2) and ECAL central cell (6), and an anti-coincidence with veto signal (3). For electrons, magnet off and radiator out, trigger is made by the coincidence of beam counter, veto counter and a high energy signal at 4 (1)
the ECAL central cell. It was possible, without apparent loss of efficiency, to set the discriminator threshold for all counters high enough to not see noise pulses. Beam counter (1) was 60 m before the detector and, to be in time to synchronize the trigger with the detector signals, it was mandatory to use a low loss fast (0.82c) coaxial cable (50 Ω,Type C-50-6-1) of 75 m long. This kind of cable was also needed to transmit the tagger signal (70 m). The SPD prototype cell (40x40x10 mm) was read with a focusing light guide and the Hamamatsu R5900U PMT, whose features (QE = 20 % at blue, gain = 2 10 6 at 800 V and signal rise time = 1.5 ns) are well suited to the needed scintillation measurements. The trigger signals (1,2,3,4) were added to the common DAQ system, allowing an off-line cross-check of the trigger conditions and efficiencies. 4 Strategy of analysis This chapter explains what was done in order to get from the raw data the distribution of deposited energies in the SPD by an electron and the deposited energies by a photon through interaction and backsplash. From these distributions, the probability of photon misidentification due to interaction and backsplash can be determined. At the time to analyze the photon and electron signals at the testbeam, three major effects not related with the SPD must be taken into account: ffl ECAL central cell is 120x120 mm and SPD is 40x40 mm. This means that ECAL trigger does not ensures us that the high energy particle really crossed the SPD: an event selection is needed. ffl The photons can produce an electron pair in the air between veto and SPD. ffl More than one photon per event can hit the SPD. For instance, the electron beam may have some photon contamination or the electron can radiate a second (third,...) photon at the radiator or at the magnet. Events where the high energy particle did not cross the SPD are called failed events. To avoid failed events, photon and electron hit position is determined computing the center of gravity of the signals from PS 3x3 central cells, cells that are as large as the SPD cell. The event selection criteria will ask that the center of gravity of the PS signals is inside a certain zone and the boundaries of this zone are chosen in order to minimize systematic errors induced by failed events in the backsplash determination. Accepting that all electrons that hit SPD are clearly detected by the SPD, when running in the electron mode, a fairly good estimation of the beam angular spread can be done and the number of failed events can be controlled. 5
Hence, these electron runs can be used to set the boundaries described above. The remaining failed event contamination is known for electron runs and can be estimated for photon runs. This estimation is done relating failed event contamination with beam angular spread and assuming that the angular spread for photons is the convolution of angular spreads coming from electron beam and multiple scattering in the radiator. The total number of entries, needed for normalizations, is the difference between number of accepted events after selection and the estimated remaining failed event contamination. Since air has a radiation length of about 300 m and there is 1.5 m of air between the veto and the SPD, about 0.5% of photons with enough energy are expected to produce an electron pair. The number of seen pairs in a set of events is proportional to the total number of photons that crossed the air between veto and SPD, hence, it can be used to estimate the number of photons per event that hit the SPD and can interact producing electron pairs. To preestimate the influence of multiphotonic events on the SPD signals, a simulation for the radiator, tagger and magnet system was done. The energy spectra of the first, second and third most energetic photon that will hit the SPD is shown in figure 5. Since the second most energetic photon has a low probability of having an energy above few hundred MeV, the energy deposition through backsplash is assumed negligible for those photons (figure 3), and backsplash is considered to be caused only by the most energetic photon. Energy deposition through interaction is considered to be caused by all photons of energy high enough to interact with the air or the SPD producing charged particles. Hence, second, third and next photons, have a non-negligible probability ofcontributing. The method used to study the photon signals accepts that, as shown in figures 1 and 4, the probabilities of having not negligible signals due to backsplash and interaction are very low and uncorrelated. Hence, eq. (2) is a good approximation for the relation between total signal (S), signal due to backsplash (B) and signal due to interaction (I): S(E) = B(E) + mi(e) = B(E) + m (I SPD + I air ); (2) where E refers to the beam energy. In this equation it is explicitly shown the consideration that backsplash is caused only by the most energetic photon, that many photons per event, m, can contribute by interaction with air (I air ) or with SPD (I SPD ) and that interaction does not depend on the beam energy. Note that m only counts those photons that can produce electron pairs and, in eq. (2), it is assumed that the number of photons that can produce an electron pair is the same for air and for the SPD. It also assumes that all pairs created in the air will be seen by the SPD. 1 1 Low energy photons, below 1-10 MeV, can deposit energy in the SPD through other processes, namely Compton scattering or photo-effects. It is considered that low energy photons are not a major problem in the designed test beam since the veto itself absorbs part of this radiation and part of its influence is removed when adjusting SPD pedestals. 6
To extract backsplash and interaction contributions to the deposited energies by a single high energy photon, photon multiplicity effects and pair production between veto and SPD must be subtracted. This will be the procedure: ffl The distribution of deposited energies in the SPD by one electron is used to obtain, by convolution with itself, the shape of the signal of two electrons. To estimate m in a data set, this two-electron signal with free amplitude, for mi air, plus an exponential, for the rest of the signal, will be fitted to the photon signals. 2 ffl In a set of data with no backsplash, m is fitted. Then, I SPD is determined from the signal and m: I SPD = S m I air (3) ffl In a set of data with backsplash, a new m 0 is fitted and B(E) is determined from the signal, m 0 and the previously determined I SPD : B(E) = S 0 (E) m 0 (I SPD + I air ) (4) This strategy has been tested using a Monte Carlo simulation. Monte Carlo information about the number of produced pairs in the air, translated properly to m, and about deposited energy due to backsplash and interaction is compared with the output of the analysis strategy. The obtained results show (table 2) that the systematic errors in the interaction and backsplash determination are, at most, at the 10% level. m(out)=m(mc) I mis (OUT)=I mis (MC) B mis (OUT)=B mis (MC) 1.02 ± 0.03 0.99 ± 0.03 1.07 ± 0.05 1.15 ± 0.13 Table 2: Results of the simulation done to test the strategy. MC input is compared with strategy output. The probability of photon misidentification is determined applying a cut at 0.7 MIPs. 2 A fit with a gamma distribution function plus an exponential, that seems to describe the data better, has also been done. Both fits give compatible results. 7
5 Results The SPD was centered with the beam maximizing the number of electrons hitting it. Electron signals were studied at 20 GeV and 50 GeV electron beam energies and photon signals were studied at 20 GeV and 50 GeV electron beam energies for three different radiator widths. The addition of an extra scintillator, between SPD and PS, provided an scenario where backsplash at SPD was negligible 3. These different runs are compiled in table 3. In figure 6 typical electron and photon SPD signals after removing pedestals are shown. Looking at the electron signal distribution, it can be seen that some electrons did not cross the SPD (peak at 0 ADC counts) and that the peak of the electron signal is at 140 ADC counts. For real data, we will say that 1 MIP signal corresponds to 140 ADC counts. In the photon signal distribution, one can see the 2 electrons peak due to pair production in the air between veto and SPD, its maximum being at around 295 ADC counts. Run Trigger Beam Energy Rad. Width Triggered Number (GeV) (mm) Events 8795 e 20 0 30020 8802 e 50 0 30000 8810 fl 20 0.5 30000 8798 fl 20 1.0 30000 8819 fl 20 1.0 30000 8825 fl 20 1.5 30000 8805 fl 50 0.5 30000 8811 fl 50 0.5 30000 8804 fl 50 1.0 30000 8812 fl 50 1.0 30000 8826 fl 50 1.5 10000 8833 fl Λ 20 0.5 30000 8832 fl Λ 50 0.5 30000 Table 3: These are the main runs used in the analysis of SPD test beam data. fl Λ means that another scintillator was put between SPD and PS in order to absorb backsplash particles. The event selection is applied (see previous section) and the remaining failedevent contamination is 1.2% for 50 GeV electrons and 2.6% for 20 GeV electrons. For photons, failed-event contamination is estimated to be between 1.3%, run at 3 The added scintillator was 1 cm thick. Backsplash particles with enough energy could cross this absorber. Since Monte Carlo shows that neglecting this backsplash introduces uncertainties in the interaction and backsplash determination below 10%, we neglect it. 8
50 GeV and 0.5 mm Pb, and 4.2%, run at 20 GeV and 1.5 mm Pb. The total number of entries is corrected subtracting these contaminations from the number of accepted events. In order to maximize statistics when applying the defined strategy, photon runs are grouped in three different sets: ffl Set 1: Runs with the extra scintillator will be used to determine the deposited energy through interaction, which does not depend on beam energy. As the parameter m already counts the number of photons that crossed the SPD and can produce an electron pair, no event selection is needed for this set. The number of entries is the number of stored events. ffl Set 2: Runs without extra scintillator at 20 GeV will be used to determine backsplash contribution to the deposited energies by photons that have energies between 14 and 18 GeV. ffl Set 3: Runs at 50 GeV will be used to determine backsplash influences when photons have energies between 44 and 48 GeV. Applying the defined strategy, atwo electrons signal plus an exponential is fitted to the tails of the photon signal (see figure 7), the parameter m is determined and interaction and backsplash results are extracted. Table 4 shows, for every set, the fitted m and the total number of events entries. Set Run Entries after m Numbers event selection 1 8832 8833 60000 1.6 ± 0.4 8798 8810 2 8819 8825 30320 3.6 ± 0.9 8804 8805 8811 3 8812 8824 29100 3.7 ± 0.9 Table 4: Runs are grouped in different sets. From set 1 interaction between photons and SPD will be analyzed, backsplash is analyzed from sets 2 and 3. The number of entries is equal to the number of stored events for set 1 and is the number of accepted events with a high energy photon hitting the SPD in sets 2 and 3. Fitted multiplicity parameter is also shown. Figure 8 shows the measured distribution of electron signals and photon signals due to interaction and due to backsplash for a 50 GeV beam energy. These distributions can be translated to probability of photon misidentification when a cut in SPD signal is applied. Applying a cut at 100 ADC counts for the SPD 9
signals (this corresponds approximately to 0.7 MIPs), ο100% electrons survive the cut and 0:8 ± 0:3 % of the photons survive the cut due to interaction with the SPD and 0:9 ± 0:6 (1:4 ± 0:6 %) of the 20 GeV(50 GeV) photons due to backsplash (see table 5). These results are compatible with Monte Carlo predictions (compare with table 1) within errors. Since statistical errors are so large, a detailed and exhaustive study of systematic errors has not been included in the results. Beam energy SPD cut I mis B mis (GeV) (MIPs) (%) (%) 20 0.70 0.8 ± 0.3 0.9 ± 0.6 50 0.70 0.8 ± 0.3 1.4 ± 0.6 20 0.35 1.1 ± 0.4 1.5 ± 0.8 50 0.35 1.1 ± 0.4 2.1 ± 0.8 Table 5: Fraction of photons (%) misidentified due to interaction (I mis ) and backsplash (B mis ) when a cut is applied in the SPD deposited energy. 1 MIP corresponds to 140 ADC counts. 6 Conclusions A description of the test beam and the analysis of its data has been done. After event selection, 21k events with electron trigger and 119k events with photon trigger were used to obtain the distribution of deposited energies at SPD by electrons and photons. The probabilities of photon misidentification have been determined applying a cut on the SPD deposited energy. A cut at 0.7 MIPs (100 % efficient for electrons) means that 0:8 ± 0:3 % of the photons are misidentified due to interaction with the SPD and 0:9 ± 0:6 (1:4 ± 0:6 %) of 14 to 18 (44 to 48) GeV photons are misidentified due to backsplash. 7 Acknowledgments We would like to thank the invaluable collaboration of the Institute for Nuclear Research (INR, Moscow) group on the preparation of the photon test beam. We also acknowledge Ivan Korolko for providing us with the simulation code and for guiding us in its use. 10
References [1] The LHCb Collaboration, LHCb: A Large Hadron Collider Beauty Experiment for Precision Measurements of CP-Violation and Rare Decays. Technical Proposal, 20th February 1998. CERN/LHCC 98-4, LHCC/P4. 11
10 Monte Carlo : SPD Deposited Energy (MeV) 1 10-1 10-2 10-3 0 1 2 3 4 5 6 7 10 1 10-1 10-2 10-3 0 2 4 6 10 1 10-1 10-2 10-3 0 2 4 6 Figure 1: Normalized distribution of deposited energies in the SPD by 10 GeV electrons and 10 GeV photons (up), by 10 GeV photons through backsplash (down-left) and by 10 GeV photons through interaction with SPD (down-right) (Monte Carlo). 12
Figure 2: Fraction (%) of 10 GeV photons (dots) and 10 GeV electrons (stars) that deposit more energy in the SPD than a certain cut (Monte Carlo). Figure 3: Probability of photon misidentification (SPD cut at 0.7 MIPs) due to interaction (dots) and due to backsplash (stars) depending on the energy of the photon (Monte Carlo). 13
Figure 4: For 10 GeV photons, energy deposited through interaction vs. energy deposited through backsplash. Low energy zone is not plotted in order to allow tails to be visible. No correlation is observed. (Monte Carlo) Figure 5: Normalized distribution of energies of the first, second and third most energetic photons that hit the SPD cell in the tagged photon test beam Monte Carlo simulation. 14
Figure 6: Distribution of photon (full) and electron (not full) SPD signals for triggered events at the test beam. ADC pedestals have been removed. 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0 100 150 200 250 300 350 400 450 500 SPD (ADC counts) Figure 7: Photon signal around the pair-creation region (points with error bars), together with results of a fit to a two-electron signal plus exponential background (left) or gamma function plus exponential background (right). All results correspond to set 3: photon trigger, 50 GeV beam energy, backsplash included. 15
Figure 8: Normalized distribution of deposited energies in the SPD by electrons and 44 GeV to 48 GeV photons (up), by these photons through backsplash (down-left) and by photons through interaction with SPD (down-right). Results obtained from test beam data. 16