Available on CMS information server CMS NOTE 22/3 The Compact Muon Solenoid Experiment CMS Note Mailing address: CMS CERN, CH-1211 GENEVA 23, Switzerland September 9, 22 The rejection of background to the H process using isolation criteria based on information from the electromagnetic calorimeter and tracker. V. Litvin, H. Newman, S. Shevchenko, N. Wisniewski California Institute of Technology, Pasadena, CA 91125, USA Abstract We present the results of a full detector simulation of different types of background to the H process. The irreducible background of prompt di-photon production and the reducible backgrounds of + jets production and QCD jet production were investigated. A special goal was to simulate for the first time a large enough sample of the QCD jet background to directly estimate the di-photon misidentification and compare it with contributions from the other backgrounds. The isolation tools are important tools in separating the signal process from the background processes. Using isolation criteria based on information from the PbWO 4 electromagnetic calorimeter and the tracker, we were able to reduce the QCD jet background to a level of 15% of the total background while keeping Higgs selection efficiency reasonably high.
1 Introduction The process H provides us with one of the most promising signals by which to search for the Higgs boson in the mass region between 9 and 15 GeV [1, 2]. We present the results of a full detector simulation of different types of background to the H process. The irreducible background of prompt di-photon production and the reducible backgrounds of + jets production and QCD jet production, where one or two jets are misidentified as photons, were investigated. The QCD jet background cross section is huge ( 1 9 pb). Therefore, previous studies [1, 2, 3] of QCD jet background have either been done at the generator level or have obtained estimates for the rate r jet at which a jet would be misidentified as a photon. Due to limited computational power, however, these studies estimated the QCD jet background by simply multiplying the rates together (factor r 2 jet ). The problem with this method is that the correlations within an event are not taken into account. In addition, the simulation was done with simplified geometry, so non-gaussian tails in the resolution have not been adeuately simulated. A special goal of this study is to simulate for the first time a large enough sample of QCD jet background to directly estimate the di-photon misidentification and compare it with contributions from the other types of background. The isolation criteria, based on information from the PbWO 4 electromagnetic calorimeter and the tracker, are important in separating the signal process from the background processes. The aim of this note is to present the successful use of isolation algorithms in selecting the isolated Higgs decay photons while suppressing the huge QCD jet background. 2 Different types of background g g Figure 1: Diagrams of irreducible background: (left) uark annihilation, (right) gluon fusion. g Figure 2: Diagram of bremsstrahlung process. 2
The search for the H signal at LHC has to consider three types of backgrounds: Prompt di-photon production from the uark annihilation and gluon fusion diagrams, which provides an intrinsic or irreducible background (see Figure 1). Prompt photon + jets production consisting of two parts: i) prompt photon + the second photon coming from the outgoing uark due to bremsstrahlung (see Figure 3) and ii) prompt photon + neutral hadron (mostly isolated π ) in a jet. The background from QCD jets, where an electromagnetic energy deposit results from the decay of neutral hadrons (especially isolated π s) in a jet. Table 1 shows the cross sections of different types of background. The threshold of the transverse momentum of the parton in the hard interaction process P hard T that was used for generation is also shown [4]. The cross sections considered do not include any K-factors, which take into account higher order QCD corrections, because these higher order corrections have not been calculated for all background processes. Background process P hard T (GeV) Cross section (pb) QCD jets 35 8 1 7 + jets 25 8 1 4 Gluon fusion 25 27 Quark annihilation 25 45 Table 1: Cross sections for different types of background. K-factors are not included. 3 Simulation tools and Monte Carlo data samples For the Higgs background study we used the following data samples: Signal: 5K H events at m H =11 GeV Background: 1 million of preselected QCD jet events 5K of preselected + jets events 5K of uark annihilation events 5K of gluon fusion events All events were generated with PYTHIA 6.152 [4] at s=14 TeV. The CTEQ 4L parton density structure functions were used in generation. The simulation of tracking in the detector was done using CMSIM version 121 (CMSIM is a full detector simulation program based on GEANT3 package) [5]. The simulated events were digitized using the reconstruction and analysis program ORCA 5 2 [6]. The digitization was done at a luminosity of 2 1 33 /cm 2 /s. Due to the huge QCD jet cross section (σ 1 8-1 9 pb), a strong preselection at the generator level is needed to simulate a sufficient number of background events in a reasonable time. The QCD cross section strongly depends on the transverse momentum of the parton in the hard interaction process P hard T. Therefore, a proper choice of this parameter is needed to accurately generate QCD jet events. We have chosen P hard T to be eual to 35 GeV. Further generator level preselection was performed as follows: we looked for events with at least one pair of any of the particles, π, e, η, η, ρ, or ω that could deposit a significant part of their energies in the electromagnetic 3
calorimeter. The transverse energies of these two particles should be greater than 37.5 GeV and 22.5 GeV (note that proposed offline CMS cuts are 4 GeV and 25 GeV respectively [2]), and the invariant mass of these two particles should be in the range 8-16 GeV, which is the range in which we are going to search for the Higgs using its two photon decay mode signature. The obtained rejection factor at generator level is 3 for QCD events, which means that 3 events need to be generated in order to produce one event that will be passed through the full detector simulation. Therefore, one million simulated and reconstructed QCD events correspond to 3 1 9 generated events. A check was done to assure that we did not lose events due to the above preselection that could fake a Higgs signal. A much looser preselection was used to simulate the + jets event sample: we looked for events with at least one of any of the particles, π, e, η, η, ρ, or ω that has a transverse energy greater than 37.5 GeV. The number of simulated and generated events are shown in Table 2 for different types of background. In the last column the integrated luminosity is shown. L intg eual to 4 pb 1 corresponds to 11 hours of data-taking at a luminosity of 2 1 33 /cm 2 /s. Background Number of (GeV) Number of L intg process generated events simulated events (pb 1 ) QCD jets 3 1 9 1M 4 + jets 3 1 6 5K 4 Gluon fusion 5K 5K 18 Quark annihilation 5K 5K 11 Table 2: Number of generated and simulated events for different types of background. 4 Isolation 4.1 Isolation based on the electromagnetic calorimeter Energy deposits from electromagnetic showers in the electromagnetic calorimeter are constructed into basic clusters using the Island clustering algorithm [7]. These basic clusters are in turn clustered into super-clusters to recover the energy radiated from photon conversions, which falls outside the seed shower cluster [7]. The reconstructed supercluster is associated with a photon candidate. The isolation criteria is based on the sum of transverse energies deposited in basic clusters in some cone R (R= δη 2 + δφ 2 ) around a photon candidate. The basic clusters that belong to the photon candidate s supercluster are not counted as part of the sum. The algorithm contains two parameters: The size of the cone R around the photon candidate wherein the transverse energies deposited in the basic clusters are summed. The transvere energy sum threshold E thresh T. If the sum of transverse energies is below this threshold, the photon candidate is considered isolated, otherwise non-isolated. The detailed results are presented in [8]. There is no strong dependence for the QCD jet rejection factor on the cone size R, though slightly better rejection factors are empirically obtained for a cone size R =.3 -.35. The cone size R =.3 is used in this study. 4.2 Isolation based on the tracker The isolation criteria is based on the number of charged tracks with p T greater than some p T threshold, p thresh T, calculated in some cone R (R= δη 2 + δφ 2 ) around the photon candidate. The algorithm contains three parameters: 4
The size of the cone R around the photon candidate, wherein the number of charged tracks is counted. The p T threshold, p thresh T. Only charged tracks with p T greater than p thresh T are considered in isolation calculations. The number of tracks threshold N thresh. If the number of charged tracks in cone R with p T greater than the chosen p thresh T is greater than N thresh, then the photon candidate is considered non-isolated, otherwise isolated. The detailed results are presented in [9]. The jet rejection factor is very sensitive to the number of tracks threshold, N thresh. By increasing from N thresh = to N thresh = 1, one can improve the Higgs signal efficiency by 6-1%, but the jet rejection factor drops by a factor of 2. Therefore, we decided to fix the parameter N thresh to be eual to zero. Slightly better rejection factors are empirically obtained for p thresh T = 1.5 GeV and for cone size R =.3 -.35. The cone size R =.3 and p thresh T = 1.5 GeV are used in this study. 5 Results Before studying photon isolation, the 12 GeV Level-1 double-isolated electron trigger [1] was applied to both the signal and background data samples. The transverse energies of the two photon candidates are reuired to be greater than 4 GeV and 25 GeV [2] respectively, and the invariant mass is reuired to be in the range 8-16 GeV. The fiducial volume in rapidity was restricted to < η < 1.4442 in the barrel and 1.566 < η < 2.5 in the endcap for both photon candidates [2, 11]. The cross section of different backgrounds in fb/gev are shown in the second column of Table 2 after the above cuts were applied. One can see that before the isolation cuts are applied, the QCD jet cross section is 25 times higher than the cross section of the irreducible background. 2 16 Gluon fusion 12 8 4 2 16 Quark annihilation 12 8 4 8 6 QCD jets 4 2 8 6 + jets 4 2 Figure 3: Invariant mass distribution of different backgrounds. Figure 2 shows the invariant mass distribution of the different backgrounds. As expected, the behavior is smooth and decreases with increasing invariant mass. The last two columns in Table 3 show the effect of the isolation algorithms. One can see that isolation is a very effective tool in reducing the huge QCD background. After isolation is applied, the QCD background is only 15% of the total background. The remaining + jets background consists of two parts: i) prompt photon + isolated bremsstrahlung photon and ii) prompt photon + isolated neutral hadron (mostly π ) in hadronic jets. Both backgrounds are 5% of the total + jet background. So, the total irreducible background, when there are two real photons in the final state is 13 fb. The remaining reducible QCD and + jets background, when π could fake a photon, is 9 fb. This latter background could be further reduced using π rejection based on lateral shower shape [2, 12]. 5
Background All cuts before Track ECAL process isolation cuts isolation isolation QCD jets 25 1. 3.2 + jets 12 19 112.4 Gluon fusion 42.3 35.7 33.5 Quark annihilation 49.7 4.9 39.1 Higgs efficiency 44.% 35.6% 33.5% Table 3: Background cross sections, dσ/dm in fb/gev, and Higgs signal efficiency. The Higgs efficiency is also shown in Table 3. The overall Higgs efficiency is 33.5% after all cuts, including the isolation cuts, were applied. 6 Acknowledgements We would like to thank J. Pino for help in writing the analysis program. References [1] C. Seez and J. Virdee, Detection of an intermediate mass Higgs boson at LHC via its two photon decay mode, CMS TN/92-56 (1992) [2] CMS Collaboration, The Electromagnetic Calorimeter Project - Technical Design Report, CERN/LHCC 97-33 (1997) [3] V. Tisserand, The Higgs to two photon decay in the ATLAS Detector, ATLAS Internal Note, PHYS-NO-9 (1996). [4] http://www.thep.lu.se/torbjorn/pythia.html [5] http://cmsdoc.cern.ch/cmsim/cmsim.html [6] http://cmsdoc.cern.ch/orca [7] E. Meschi et al., Electron Reconstruction in the CMS Electromagnetic Calorimeter, CMS Note 21/34 (21) [8] V. Litvin, H. Newman, S. Shevchenko, Isolation studies for electrons and photons based on ECAL, CMS IN 22/23 (22) [9] V. Litvin, H. Newman, S. Shevchenko, N. Wisniewski, Isolation studies for photons from Higgs decay based on the Tracker, CMS IN 22/34, (22) [1] CMS Collaboration, The Level-1 Trigger - Technical Design Report, CERN/LHCC 2-38 (2) [11] C. Seez, The effect of tracker material on the Level-1 electron trigger, CMS Note 2/31 (2) [12] L. Borissov, A. Kirkby, H. Newman, S. Shevchenko Neutral pion rejection in the CMS PbWO 4 crystal calorimeter using a neural network, CMS Note 1997/5, (1997) 6