Available on CMS information server CMS NOTE 22/? The Compact Muon Solenoid Experiment CMS Note Mailing address: CMS CERN, CH-2 GENEVA 23, Switzerland xxxx Study of a High Level b-trigger selection of H fully hadronic decays D. Benedetti, L. Fanò INFN and University of Perugia, Perugia, Italy Abstract We present the High Level Trigger selection of the channel decaying into hadronic final states at Low Luminosity. The basic idea is to use a fast b-tag where b jets are identified with a simple algorithm after the Level selection. The main source of background considered is direct QCD multijet production. We summarize the efficiency on signal selection and background rejection and finally we give an estimate of the expected event rate.
Introduction One of the most promising channels in which a low mass Higgs search could be successfull is the associated production where the Higgs decays into a pair of b quarks. The reconstruction of two top quarks in the final state allows the background to be suppressed. At trigger level the signal is well detected by means of a high lepton, coming from one of the top decays, together with a large jet multiplicity. The non-leptonic decays represent a larger fraction of the top decays, but the huge QCD background at trigger level discourage their subsequent use. Considering the H in the semileptonic decay mode (29% of the total branching ratio) it is possible to achieve for example, with 3 fb, a signal significance of 5.3 for a Higgs mass of 5 GeV/c [] in the Standard Model scenario. It is also possible to measure the top Higgs Yukawa coupling using the event rate. A high discovery potential in the Minimal Supersymmetric extension of the SM could also be investigated in the maximal m scenario[]. In this case with 6 fb of integrated luminosity, most of the tan - m available parameter space can be covered. For the fully hadronic final state (46% of total branching ratio) it is demontrated with fast simulation studies[2] that a signal significance of 3.5 can be reached supposing, as for sempileptonic case, a simulated energetic L trigger requiring at least 8 jets in the final state with E 2 GeV in 2.5. However, the simulated trigger is not realistic for two reasons: ) the thresholds considered are low[3] and 2) % efficiency is assumed. Also under this condition the collected signal significance is not so high: an improvement is needed for the signal selection. In this study a high level trigger strategy is introduced and its effect on H channel in a fully hadronic final state analysed. Figure : The event in the fully hadronic final state In the case of event reconstruction on fully simulated and reconstructed events, not discussed here, the hadronic final state (Fig. ) presents two main difficulties: firstly that the two b quarks coming from Higgs decay are soft and secondly that eight jets in the final state is a very complex system to reconstruct. The relevant cross section for the signal and the main QCD interval used as background are listed in Table. Table : Cross sections and expected rates for signal and QCD analysed processes Process Cross Section expected low luminosity rate (2x cm.36 pb 2x Hz 5 GeV) 577 mb 4x Hz QCD (3 QCD (5! 8 GeV).2587 mb 4x" Hz QCD (8 QCD (2 # 2 GeV) 3.6294 $ b 7.2x Hz 7 GeV) 9 $ b.2x Hz s ) Only QCD multijets direct production has been considered: other resonant and non resonant backgrounds like Z, and are expected to contribute according to Table 2. 2
Table 2: Cross sections for other background contributions to the fully hadronic final state. Process Cross Section expected low luminosity rate (2x cm s ) 9 pb 8x Hz.5 pb 3x Hz 233 pb 4.66x HZ 2 Samples description This study was carried out using events in fully hadronic final state. A total of 6 events were generated and events fully digitised and simulated using the official CMS 22 production. The generation has been done using Pythia[4] version 6.58, the simulation using cmsim[5] 25 and the hit formatting and reconstruction steps using ORCA 6 [6]. The generator switches for production and decay channels used are the following: MSUB(2)=: gg H MSUB(22)=: H MDME(24,)=: H MDME(9-25,)=: all hadronic W decay switched on. A simple investigation has been done on the sample at the generation level to check the sample quality and test the reconstruction strategy that is intended to be adopted. In this case the subroutine [4], a UA-like algorithm, has been used at generator level to define the jets using a cone of R =. Events have been preselected requiring at least eight jets in the final state with E GeV. Eight jets are necessary to reconstruct the whole event as will be shown. The efficiency on event selection, in this case, is more than 3%. 4 2 ID Entries Mean RMS 22 5994 9.245.925 25 2 ID Entries Mean RMS 2 5994 3.7 256 8 5 6 4 2 5 2 4 6 8 2 4 6 2 4 6 8 2 4 6 Figure 2: )Histogram of jets per event. 2)Histogram of b partons associated with a jet per event. Figure 2 shows the number of reconstructed jets per event (left) and the number of jets identified as coming from b quarks (right). A jet is flagged to come from the hadronisation of a b quark if the R between the jet and quark direction is smaller than.3. If more than one jet is matched, then the nearest one is taken. The mean number of reconstructed jets per event is 9.2 while on average 3. jets are associated with a b parton. The W bosons have been reconstructed (without any calibration!#" jet energy) &" taking.- the best combination 2!#" &" of 4 nonb jets that gives the best value for a combined = %$ (' *),+ / $ 3',)*+ variable with the W mass set to 8 GeV/c (Fig.3). The W boson mass is reconstructed with 76.84 9. GeV. After the two W bosons have been reconstructed, the top quarks are searched for, using two out of the four b jets in the final state (fig.3); also in this case the correct jets are chosen to give the best combination for the top identification with a supposed mass of 75 GeV/c. The reconstructed mass for the top is 68.4 2 GeV. Finally the two remaining b jets can be used to reconstruct the Higgs boson mass. 3
Figure 3: ) Reconstructed W mass 2) Reconstructed top mass The QCD samples used are datasets of the official cms production 22[7]. The total number of QCD events used is about 7. 3 The High Level b-trigger In order to speed up the reconstruction, only tracks within a jet cone will be used. The performance of the tagger depends crucially on the quality of the tracks and the jet direction. Tracks are reconstructed within the jet defined by the direction given by the Level- calorimeter jet candidate. The cone size is decided as a function of the reconstructed jet and its apex given by the reconstructed primary vertex of the hard interaction. The primary vertex location along the beam line is reconstructed using the algorithm presented in reference[8], which makes use of the so-called pixel-lines. Track reconstruction is based on the partial reconstruction of tracks using a regional approach: starting from the pixel lines as seeds, subsequent hits are sought in a region compatible with the track in a region around the jet axis using a Kalman Filter technique. The reconstruction is stopped after a suitable number of hits are found along the trajectory. σ(d rec-d sim)[µm] 8 7 σ(p t rec-p t sim)[gev/c] 5 6 5 4. 3 2 4 6 8 2 Reconstructed Hits.5 2 4 6 8 2 Reconstructed Hits Figure 4: Distribution of the impact parameter (left) and momentum resolution (right) for partial track reconstruction, compared with full track reconstruction (leftmost point) as a function of the number of hits. Figure 4 shows the error on the track and transverse impact parameter as a function of the number of hits along the track, compared with the full tracker resolution. At about 5 hits the precision comparable with the full reconstruction. 4
A smaller number of hits, however, increases the fraction of ghost tracks. The fake rate decreases to below the % level for tracks with at least 5 hits. Tracks that have been reconstructed within the jet cone can be used to better determine the Level- jet direction. In fact, due to the large granularity of the Calorimeter trigger cells, the resolution of the Level- jets is rather poor, both in as well as in. The effect of a reduced angular resolution of the jet direction causes a sign flip of the track impact parameter, causing a worse performance of the tagger. The new direction is determined by a weighted sum of the tracks and and gives, on average, a factor of two improvement in resolution. 4 Triggers Results All the results are reffered to the non-staged scenario for the CMS detector. In this section the effect of three different trigger strategies on signal selection and background rejection will be analysed. The first strategy is the Level trigger based on jet energy threshold. With the second strategy the effect of final states containing b-jets is analysed. The last strategy describes how combining the Level trigger with the developed fast b-tag can give the best performance. 4. Level- Trigger Fig. 5 shows the effect of an energetic treshold trigger on the signal selection efficiency and background rate. The transverse jet energy has been calibrated with the formula[9][]: &" E!#" &" 2!#" &" = a E %$ + b E $ +c!" &" Where the coefficients a,b and c are function of L transverse energy (E $ ) and. The correction used is valid up to 5. jets 2 jets 3 jets 4 jets 7 6 5 4 3 2 LOW LUMINOSITY jet 2jet 3jet 4jet 25 5 75 25 5 75 2 225 25 Calibrated jet E T (GeV) 5 5 2 25 3 Calibrated jet E t (GeV) Figure 5: The first and second plot show respectively the signal selection efficiency and the qcd rate as a function of the calibrated transverse energy cut. In both cases rates for, 2, 3 and 4 jets in the final state have been plotted The jets are preselected with a cut on reconstructed energy of GeV corresponding to 95% of signal event and 5x" Hz of 4-jets QCD rate. The Level- jet selection could be performed by requiring at least four jets in the final state. A total of 8% efficiency on the signal events is &" achieved at the expenses of about KHz rate by requiring the four jets to exceed at least 4 GeV in the E. If instead the energy requirement is rised to 5 GeV the signal selection efficiency drops to 65% with a QCD rate of 2 Hz. 4.2 The HL-bT In this section we will show how a b-jet identification could lower the rate while keeping a sufficiently large efficiency for the signal. The b-tag algorithm chosen is the track counting method in two dimensions []. As stated in section 3 the tracks have been reconstructed by using a conditional reconstruction. In this paper we have stopped the track 5
reconstruction when 5 hits are reached, and compared the results with the results with a reconstruction which makes use of a maximum of hits along the track. The trigger strategy is now applied to select different signal samples as a function of the b-tag request. Jets are reconstructed with a R smaller than, and the original Level- direction recomputed using the reconstructed tracks as explained in Section 3. Jets are b-tagged by the request of at least two tracks with impact parameter significance greater than different values: from to 5, for tagged jet, and from to 2.5 (with steps) for 2, 3 or 4 tagged jets. The results are shown in fig.6. These results (circle) have been compared with the hits (box) reconstruction scenario..3.3. hit. hit.3. 2 3 4 5 2 3 4 5 hit.3. 2 3 4 5 2 3 4 5 hit Figure 6: Signal selection versus QCD rate after b-trigger request. Different plots are for,2,3 or 4 tagged jets using partial (box) and full (circle) track reconstruction. Jets are preselected with E GeV Also in this case the jets are preselected with E GeV. Now the effect of different signal efficiencies requests can be studied. A 5% efficiency corresponds to a QCD rate of 2 KHz for b-tagged jet. The QCD rate can be further decreased requiring, for example, 3 tagged jets: in this case a 5 Hz rate with a signal efficiency of 3% can be obtained. 4.3 Combined Triggers The second applied strategy is to combine two different triggers. Fig. 7 shows the results of using both the energetic and the b-trigger criteria: jets are selected with calibrated E 5 GeV and, separately, for, 2, 3 or 4 tagged jets. In this case it is possible, by requiring one tagged jet, to select the 55% of the signal events and reduce the Low Luminosity QCD rate to 2 Hz. In table 3 are summarized signal efficiencies and QCD rates for analysed triggers. 6
AT LEAST 4 JETS E T 5(Gev) AT LEAST b-tagged JET AT LEAST 4 JETS E T 5(Gev) AT LEAST 2 b-tagged JET.3.3. hit. hit 2 3 2 3 AT LEAST 4 JETS E T 5(Gev) AT LEAST 3 b-tagged JET AT LEAST 4 JETS E T 5(Gev) AT LEAST 4 b-tagged JET.3.3. hit. hit 2 3 2 3 Figure 7: Signal selection versus QCD rate after b-trigger request. Different plots are for,2,3 or 4 tagged jets using partial (circle) and full (box) track reconstruction. Jets are preselected with E 5 GeV Table 3: Summary of signal efficiency and QCD rates for different trigger criteria. Trigger type Signal Efficiency QCD rate 4 jets with E GeV 95% 5x" Hz 4 jets with E 5 GeV 65% 2 Hz 4 jets with E GeV + b-tagged jet (3 b-tagged jets) 5% (3%) 2x Hz (5 Hz) 4 jets with E! 5 GeV + b-tagged jet 55% 2 Hz 4.4 Signal Estimation The event rate for signal and QCD background can be calculated explicitly over the relevant range for the Low Luminosity LHC scenario. Table 4 summarises the number of events expected for 3 of integrated luminosity for different trigger criteria. 5 Conclusions and perspectives In this work a high level trigger based on b quark selection has been deeply analysed. In particular the Higgs boson with associated production in a fully hadronic final state has been investigated. It has been shown that the QCD background in the low luminosity scenario can be reduced from 5x" Hz to 2 Hz while selecting the 55% of the signal events. This is done by requiring 4 jets in the final state events with E 5 GeV coincident with at least one b-triggered jet. The next steps will be to inspect the full signal reconstruction capabilities considering all background sources and study the High Luminosity scenario. 7
Table 4: Signal and QCD events for different trigger criteria in 3 of integrated luminosity Trigger type # Signal events # QCD events 4 jets with E GeV 26 7.5x 4 jets with E 5 GeV 72 3x 4 jets with E! GeV + b-tagged jet (3 b-tagged jets) 54 (324) 3x (7.5x ) 4 jets with E 5 GeV + b-tagged jet 594 3x References [] CMS NOTE 2/54 - Searching fro Higgs Bosons in Association with Top Quark Pairs in the Decay Mode - V. Drollinger and Th. Muller [2] PhD Thesis - Reconstruction and Analysis Methods for Searches of Higgs Bosons in the Decay Mode at Hadron Colliders - V. Drollinger [3] L Trigger Tables - http://computing.fnal.gov/cms/jpg/level table 2.jpg [4] CERN-TH.72/93 - Pythia and Jetset Physics and manual - T. Sjostrand [5] The CMSIM Home Page - http://cmsdoc.cern.ch/cmsim/cmsim.html [6] The ORCA Home Page - http://cmsdoc.cern.ch/orca/orca.html [7] The CMS Production Page - http://cmsdoc.cern.ch/cms/production/www/html/general/indexsql.html [8] CMS-CR-22-3 - High Level Tracker Triggers for CMS - D. Kotlinski, A. Starodumov [9] CMS IN 2/ - Energy Corrections for QCD Jets - S. Abdullin, S. Arcelli et al. [] Updated Correction Routine - http://computing.fnal.gov/cms/jpg/spring22/ [] CMS NOTE 22/ G. Segneri and F. Palla, Lifetime based b-tagging with CMS. 8