KEK preprint 94-180 NWU-HEP 95-01 DPNU-95-02 Electron identification using the TOPAZ detector ∗ 5 at TRISTAN 9 9 1 Masako Iwasakia†, Eiichi Nakanob‡, and Ryoji Enomotoc§ n a J aDepartment of Physics, Nara Women’s University, Nara 630, Japan 1 bDepartment of Physics, Nagoya University, Nagoya 164, Japan 1 cNational Laboratory for High Energy Physics, KEK, Ibaraki 305, Japan 1 v 4 0 Abstract 0 1 We present an electron-identification method using the time-projection chamber and the lead- 0 glass calorimeter in theTOPAZdetectorsystem. Usingthismethodwehaveachieved good electron 5 identification against hadron backgroundsover a wide momentum range in the hadronic eventspro- 9 ducedbybothsingle-photonexchangeandtwo-photonprocesses. Pion-rejection factorsandelectron x/ efficiencies were 163 and 68.4% for high-PT electrons and 137 and 42.7% for low-PT electrons in the single-photon-exchangeprocess, and 8600 and 36.0% for thetwo-photon process, respectively. e - p e 1 Introduction h : v The TOPAZ detector is located at the TRISTAN e+e− collider of KEK. The center-of-mass energy was Xi 58 GeV and the integrated luminosity will be more than 300pb−1 by the end of 1994. Electron identification at high purity and high acceptance is necessary to tag heavy-quark events. r a In the e+e− annihilation process, the forward-backward asymmetry of quark-pair production becomes maximum at the TRISTAN energy region. A high-accuracy measurement of this was achieved [1, 2]. In two-photon processes, open-charm production was also studied with high accuracy [3]. Especially for low-P charm production, electron identification provides the best prove for resolved-photon processes T [4], as well as for higher order corrections of the QCD [5]. Inordertocarryoutthese studiesweneedgoodelectronidentificationoverawidemomentumrange. In particular, the detection of low-momentum electrons is necessary. To meet these requirements we have developed a method which uses an energy-loss measurement with a time-projection chamber and an energy measurement with a lead-glass calorimeter. In this article we present related details. 2 The TOPAZ detector system The TOPAZ detector is a general-purpose 4π spectrometer featuring a time-projection chamber (TPC) as its central tracking device [6, 7, 8, 9]. A schematic view is shown in Figure 1. Charged tracks were ∗submittedforpublication. †internetaddress: [email protected] ‡internetaddress: [email protected] §internetaddress: [email protected] 1 MDC BCL BDC Pole Base Super Conducting Coil TOF RCAL Pole Tip T P C TPC E ECL S D e c C t o r TCH VTX B QCS G BGO O Beam Pipe FCAL LUM Figure 1: Schematic view of the TOPAZ detector. 2 Components Material and size Measured Performance Magnet 2.9mφ 5m, 0.7 X0 B=1.0 T × Beam pipe 10cmφ, 1mmt Be Vertex Chamber Jet-chamber σ=40µm TCH Cylindrical Drift Chamber σ=200µm TPC 2.4mφ 2.2m σ =185µm rφ × 16 sectors σ =335µm z 10 pad rows, 176 wires / sector σ =4.6% dE/dx Ar/CH4 (90/10)3.5atm σPT/PT = (1.5PT)2+1.62% TOF Plastic Scintillator σ=220ps p BDC Streamer tube σ=350µm BCL Lead-glass (SF6W), 20X0 σE/E = (8/√E)2+2.52% 4300 blocks σθ=0.38oq cosθ 0.82 | |≤ ECL Pb-PWC sandwich σ /E =6.7% E 18 X0 (for 26-GeV electron) 0.85 cosθ 0.98 σ =0.7o θ ≥| |≤ Table 1: Performance of the TOPAZ detector components. detected with the TPC placed in an axial magnetic field of 1.0 tesla. The energies of the photons and electrons were measured using a barrel lead-glass calorimeter (BCL) and an end-cap Pb-PWC sandwich calorimeter (ECL). A summary of the TOPAZ detector is given in Table 1. The coordinate system used in this article is as follows: z is defined as the beam-electron direction; x is the vertical axis and y is the horizontal axis. The major components of the TOPAZ detector are described in the following sections. 2.1 The Time-Projection Chamber (TPC) TheTime-ProjectionChamber(TPC)isathree-dimensionaltrackingchamberwhichcanidentifyparticle speciesbyenergy-lossmeasurements. Ithasafiducialvolumeof2.4mindiameterand2.2minaxiallength, comprises 16 multiwire proportional counters (sectors), each equipped with 175 sense wires and 10 rows of segmented cathode pads, and is filled with a gas mixture of 90% Ar and 10% CH4 at 3.5 atm [6]. ChargedparticlespassingthroughtheTPCsensitiveregionionizegasmoleculesandliberateelectrons alongtheirtrajectories. Theseelectronsdriftalongthedirectionoftheelectricfieldtowardthesectorsat a speed of 5.3 cm/µsec, and finally produce avalanches around sense wires; induced signals are detected by cathode pads. In order to reduce any space-charge effects, the gating-grid plane is placed above the shielding-grid plane. The signals from wires and pads are amplified and are shaped by analogue electronics. They are then fed into charge coupled device (CCD) digitizers sampling the wave-forms in unitsofa100nsec“bucket”fromwhichthez-positionsandenergy-loss(dE/dx)informationareextracted. The spatial and momentum resolutions are studied with cosmic rays as well as Bhabha and µ+µ− events. Thespatialresolutionwasobtainedtobeσ =185µmandσ =335µm. Themomentumresolution xy z was obtained by comparing two measurements in the opposing sectors in cosmic-ray events to be σ /P = (1.5P )2+1.62%. PT T T p The dE/dx resolution was studied using minimum-ionizing pions in the beam events. The obtained resolution is σ =4.6% after calibration (discussed later). dE/dx 3 TOPAZ detector h - e sector + e x y z x z Figure 2: Definition of the coordinate system in the TOPAZ detector. 2.2 The Barrel Lead-Glass Calorimeter (BCL) The Barrel Calorimeter (BCL), which comprises 4300 lead-glass Cˇerenkov counters, has a cylindrical structure covering an angular region of cosθ 0.82. The calorimeter is divided into 72 modules, 8 in | | ≤ the φ direction and 9 in the z direction, each of which has an array of 60 lead-glass counters [7]. Eachlead-glasscounterismadeofSF6Wandhas20radiationlengths,sothatitcanabsorbmorethan 95% of the shower energies at TRISTAN’s highest energy. The lead-glass blocks are tilted by 1.8o with respect to the radial line in the r φ plane so that no photons coming from the interaction point escape through the counter-to-counter ga−ps. The Cˇerenkov light emitted in a lead-glass counter is detected by a photomultiplier after passing through a light guide. The signal from the photomultiplier is sent to a digitizer and is used for a neutral trigger [9]. The system is calibrated by using a Xe flash lamp. The performance of the BCL has been studied by using Bhabha events. The energy and spatial resolution were determined to be σ /E = (8/√E)2+2.52% and σ =0.38o, respectively. E θ q 3 Track reconstruction Therawdatawasfirstcorrectedforanychannel-to-channelvariationofelectronics. Thespatialpositions of “hits” in the TPC were then determined using pad signals, and particle tracks were searched among theses points. Each found track was fitted to a helix, and the charge and momentum were assigned. We define the coordinatesysteminthe TOPAZdetectorasshowninFigure 2. The globalcoordinate in the TOPAZ detector is defined by x, y and z. The origin of the coordinates is the detector center. The ξ, η and ζ are coordinatesfixed onthe TPC sector. The ξ-axis is alongthe sense-wire direction,the η-axispoints in the radialdirectionatthe center line ofthe sector,andthe ζ-axis is alongthe beamline. The details concerning this section can be found in reference [10]. 4 3.1 Space-point reconstruction The pad signal of each channel was first corrected for any electronics variation, and was then examined asa functionofthe CCD bucketnumber. Contiguoussignalsweregroupedinto acluster to forma ζ hit. The peak pulse height and the ζ position of the cluster were calculatedby fitting the highest three CCD buckets to a parabola. Clusters were then searched in the ξ direction. If signals in contiguous pads had roughly the same ζ positions (within 1cm), theses signals were combined into a cluster. When a cluster was found, the ξ and η positions of the cluster were determined. Since the charge distribution along the pad row has approximately a Gaussian shape, the ξ position was calculated by fitting the highest 3 pad pulse heights to a Gaussian. If the cluster had only two pad signals, the pre- determined width of the Gaussian was used in the fit. The η position of the cluster was calculated by taking the centroid of the signals of five wires nearest to the pad row. Thepositionresolutionsinξ andζ werestudiedusingcosmic-raytracks;typicalvalueswereobtained as σ =230µm and σ =340µm. ξ ζ 3.2 Track reconstruction To find tracks from the reconstructed hits in the TPC, all found space points were first filled in a 2- dimensional histogram with φ and z(R ) axes such that ref φ = tan−1(x/ x2+y2) z(Rref) = z (Rrefp/R), × whereR istheradialpositionofthefixedreferencepoint. The spacepointsassociatedtoatrackfrom ref the originwith arelativelyhighmomentumwere clusterizedina bin ofthe histogram. The trackfinding beganby searchingfor a bin containingmore spacepoints thana giventhreshold. The thresholdwas set at 9 in the beginning, which correspondedto 10 padraw hits; it was then loweredby one when allof the bins above the currentthreshold were found. Loweringof the threshold was repeated until the threshold became two (three pad rows). To test whether the space points in the bin formed a track or not, three space points which spanned the largest lever arm were chosen and checked as to whether they were on a single helix. If the space pointswereonahelix,otherspacepointsinthebinwerecheckedastowhethertheywereconsistentwith the helix. If not, another set of three space points were chosen and the test was repeated. If a sufficient number of space points could be associated they were recognized to form a track and were fitted to a helix function. Before the helix fitting, the space points were correctedfor any dependence on the track angle to the sectors. The helix function was parametrized as follows: 1 x= (cosφ0 cos(φ+φ0))+drcosφ0+X0 κ − 1 y = (sinφ0 sin(φ+φ0))+drsinφ0+Y0 κ − and 1 z = ( )φtanλ+dz +Z0, − κ where (X0,Y0,Z0) is the position of the pivot, κ is the inverse of the radius of the helix, 5 φ0 is the angle of the pivot to the x axis, d and d are the distance of the pivot from the true trajectory, and r z λ is the the angle which the track makes with the xy plane. The parameters to be determined by the fit are dr,φ0,κ,dz, and tanλ. The fitting of the helix was achieved by minimizing χ2 defined as ξ ξ(η ) z z(η ) χ2 = (( i− i )2+( i− i )2), σ σ i ηi zi X where ξ and z are the measured position in the i’th pad row and η(η ) and z(η ) are the expected i i i i position by the helix. We checkedthis χ2 and made a final decisionas to whether to recognize the space points as a track or not. After all histogram bins forming tracks were found, the space points which were not recognized as tracks were then examined. From the remaining histogram bins, the bin containing the largest number of space points was searched for. The inner-most space point in the bin was chosen to be the origin (pivot) of the new histogram. The new histogram was defined with wider bins so that the space points of a low-momentum track could be clusterized in a bin as a new track candidate. The bins were tested in the same way as described above. 3.3 Track Refinement Although the gating grid prevents the feed-back of positive ions into the drift space and reduces the distortion in the electric field in the TPC, a small amount of distortion still remains, resulting in a systematicshiftinthespace-pointdetermination. ThedielectricmaterialintheTPC,suchasfieldcages, are also a source of distortion. In addition to electrostatic distortion, the spatial positions of the TPC sectors may deviate from the designed position. This also causes a shift in the determination of the space points. The electrostatic distortions in the rφ plane were parametrized as ∆xφ =a0(r rref)2, − where ∆x is the position displacement in φ due to the distortion. The deviations of the sector position φ in the rφ plane (∆ζ, ∆η) are parametrized as ∆ξ =d δη ξ − and ∆η =d +δξ. η The shift in the z direction (dz) was expressed so as to include both the electrostatic distortion and the position displacements of the sectors, and was parametrized as ∆z =d +d z. z v These parameters were defined for each sector. The parameters of the electrostatic distortions were determined based on a comparison of two cosmic-ray data taken when the beam was on and off. The parameters of the position displacements of the sectors and the shift in the z-direction were determined 6 based on a comparison of two measurements for a cosmic-ray track by two sectors when the beam was on. The space points calculated in the previous sections were corrected for any distortions using these parameters. Finally, the space points were fitted to a helix again in order to assign the sign of the charge and momentum to the track. The momentum in the xy plane (P ) and along the z-axis(P ) were expressed T L using the helix parameters as Q P = T |ακ| and P =P tanλ, L T where 1 α= =333.56cmGeV−1. cB Qisthechargeoftheparticle,c thevelocityoflight,andB the strengthofthemagneticfield(1.0tesla). The sign of the charge was determined from the sign of κ. The resolution of the momentum measurement was estimated using the cosmic rays [11]. The res- olution was studied by comparing two measurements in opposite sectors for a cosmic-ray track. The momentum resolution was studied as a function of momentum, and was obtained as ∆P T = [(1.5 0.1)P ]2+(1.6 0.3)2%. T P ± ± T p The effect of distortion corrections was studied using Bhabha events. The momentum resolution for the Bhabha events was measured to be (1.7 0.1)P %. T ± 4 Energy-loss measurement in the TPC Theenergylossperunitlength(dE/dx)isafunctionofthevelocity(β)andthe charge(Q)ofaparticle. In this section we examine the property of the dE/dx curve and discuss what determines the dE/dx resolution. In practice, the dE/dx resolution suffers from various systematic shifts due to temperature, pressure, high voltage, electronics and so on. We need to improve the resolution as much as possible in orderto maximize the searchableregion. A detaileddescriptionof the derivationof dE/dx is givenhere. 4.1 Theoretical consideration 4.1.1 Average energy loss When a charged particle traverses a medium it interacts with atoms in the medium. The dominant process of energy loss is atomic excitation. The probability of exciting an atom scales as the square of the transverse component of the incident particle’s electric field with respect to its velocity vector. The transition probability, and hence the energy loss, scales as the charge of the incident particle squared. Qualitatively, the average energy loss (<dE/dx>) in the material can be described as a function of the velocity (β) and charge (Q) of the particle. The function can be approximately written as dE 4πnQ2e2 2m β2γ2 < > (ln e 2β2), dx ≈ m e2β2 I2+(h¯ω)2β2γ2 − e where I is the logarithmic mean atomic ionization potential, 7 ω is the plasma frequency of the medium, n is the electron density of the medium, m is the electron mass, e e is the electron charge, and Q is the charge of the incident particle. Thevaluesof<dE/dx>fallas1/β2withincreasingthemomentumoftheparticles. Thefalloffisdueto the fact that a particle traversingthe materialspends less time in the electric field of atoms, and, hence, transferslessenergyastheparticle’svelocityincreases. Theionizationbecomesminimumatβγ 3,and ≃ then < dE/dx > rises logarithmically as the momentum increases. This “relativistic rise” is the result of the relativistic increase in the transverse electric field of the incident particle. The <dE/dx> finally reached the plateau due to the “density effect” caused by the polarization of the medium. 4.1.2 Energy-loss distribution There are two contributions to the shape of the energy-loss distribution: atomic excitations and δ − ray production. The δ ray contribution falls inversely proportional to the energy-transfer squared (1/(∆E)2), because this−is the classical Rutherford scattering for free electrons. This process makes a long tail, called Landau tail, in the energy-loss distribution up to the kinematical limit. The peak of energy loss distribution is due to atomic excitation, which is the dominant energy-loss process. Argon, which is the main component of the TPC gas, has three shell structures: K, L, and M. The energy levels are 3.20 keV, 248 eV, and 16 - 52 eV, respectively. Since the collision probability with an electron is roughly proportional to its energy level, the lower energy-level excitations give a narrower dE/dx distribution. 4.2 dE/dx measurement 4.2.1 Derivation of the energy loss Theenergylosswasmeasuredusingwiresignalsofthe TPC.First,the wiresignalclustersweresearched in the z direction in the same way as the pad clusters. The pulse height of a cluster was determined fromthe peakamplitude ofaparabolafittedto the highest3CCDbucketsinthe cluster. Thez position of the cluster was also determined from the fit. The cluster was then associated to a track using the z information. If the cluster was within 1 cm of the track in the z direction, and not associated to other tracks, the cluster was linked. The pulse-height amplitude was converted into the energy-loss value by two steps. The first was a correction for any non-linearity of the electronics. The second was a correction for the wire-to-wire gain variation. Details concerning the calibration are described in the following subsection. Since a TPC sector has 175sense wires, the energy loss for a track wassampled every 4-mmsegment up to 175 independent data. A single variable, which is similar to the most probable energy loss, was deduced from the data sample. The final dE/dx was then derived by making various corrections, which are discussed later. 4.2.2 Calibrations Prior to the experiment, after the electronics were calibrated by pulsing the shielding grid wires with various test-pulse amplitudes, the calibration curve for each channel was determined. The linearity of our electronics is better than 0.5% below the saturation point. The remaining non-linearity, appreciable near to the saturation point, is corrected using this calibration curve. 8 s 70000 t i h f o 60000 r e b m 50000 u N 40000 30000 20000 10000 0 0.1 0.15 0.2 0.25 0.3 Energy Loss (KeV/cm) Figure 3: Pulse-height distribution obtained by the TPC wire signals for minimum ionizing pions. Afteranelectronicscalibration,acorrectionforthewire-to-wiregainvariationismade. Eachsectoris equippedwiththreerodshavinga55FeX-raysourceforeachwire. Eachrodcanbemovedpneumatically behind the hole on the cathode plane to irradiate wires. The wire-to-wire gain variation is corrected by calibratingthe pulse height for the main peak of5.9 keVfor eachwire. This calibrationreduces the gain variationovera sector within a 3%level, and alsogives the factor necessaryto obtainthe absolute value of the measured dE/dx. 4.2.3 The 65%-truncated mean Figure3showsthemeasuredenergy-lossdistributionintheTPCforminimumionizingpions. Theshape isasymmetricandhasthe longtailatthehigherside. Therefore,asimplemeanofthedE/dxsamplefor a track is not a good parameter for particle identification. The most probable energy loss or the mean valuearoundthe peakisabetter parameter. We adaptedthe meanofthelowest65%inthe datasample (the 65%-truncated mean) as an estimator for the energy loss of a track. In the following we use dE/dx as in this meaning. 4.2.4 Corrections To achieve a good resolutionin dE/dx measurements, one must reduce the systematic shifts of the gain. In this section we discuss the systematic shifts of dE/dx and how we correct them. In the TPC, measuring the energy loss is an indirect process. An implicit assumption is that the numberofionizationelectronsproducedpertracklengthisproportionaltodE/dxofthechargedparticle. As electrons in the gas drift to the sectors they defuse; some fraction is absorbed by impurities, such as O2, in the gas. Since the maximum drift length is more than 1 m, this effect is not negligible. We determined the attenuation factor using minimum ionizing pions in the momentum range between 0.5 and0.6GeV/c. Figure4showsthemeasuredvalueofdE/dxasafunctionofthedriftlength. Wecorrect this effect using following function: PHcorrect =PHmeasured (1+C1L), × where 9 5 ) m c 4.8 / V e 4.6 K ( 4.4 x d E/ 4.2 d 4 3.8 3.6 3.4 3.2 3 20 40 60 80 100 120 Drift Length (cm) Figure 4: dE/dx as a function of the drift length. PH is the pulse height, and L is the drift length. The attenuation factor (C1) is typically 7% per meter. Thegainshiftduetothesamplethicknessdependencewasalsocorrectedusingtheminimumionizing pions. Figure 5 shows dE/dx vs. the path lengthon the logarithmscale. This systematics was corrected using PHcorrect =PHmeasured (1 C2 log(X)), × − × where X is the path length normalized by 4 mm. The factor C2 was determined to be 0.2. The electric fields near to the sense wires also affect the avalanche process. The dependence of the pulse height on the high voltage was parametrized as δV PHcorrect =PHmeasured (1 C3 ), × − × V0 where V0 is the nominal high voltage (1970V) on the sense wire, and δV is the variation of the voltage. The correction factor (C3) of the pulse height was determined using data taken in low high-voltage cosmic-ray runs. A 1% high-voltage change corresponds to a 17% change in the gain. The electrons approaching the sense wires pass through the avalanche process. As the gas density increases the mean-free path of electrons becomes shorter and the soft collisions in which the electrons do not ionize the atoms increase; hence, the amplification factor drops. The gas density is proportional to the pressure and inversely proportionalto the temperature. The pressure was well controlled, and its 10