Search for a neutral CP-even Higgs boson decaying to a Z boson and a Standard Model-like Higgs boson using tau final states.

1 2 3 4 5 Search for a neutral CP-even Higgs boson decaying to a Z boson and a Standard Model-like Higgs boson using tau final states. Stephane Cooperstein Supervisor: Professor Sridhara Dasu May 13, 2014 6 7 8 9 10 11 Abstract A search for A Zh llττ using data collected by the Compact Muon Solenoid Experiment corresponding to an integrated luminosity of 19.6 fb 1 at s = 8 TeV. No evidence is found. An upper limit on the pp A cross section times branching fraction is set in the region 260 GeV m A 340 GeV at 95% confidence level. 12 13 14 15 16 17 18 19 20 21 Contents 1 Introduction 2 2 The CMS Detector 5 3 Data and Monte Carlo Samples 6 4 Trigger Selection 6 5 Particle Identification 9 5.1 Electron Selection......................... 9 5.1.1 Loose Electron Identification............... 9 5.1.2 Tight Electron Identification............... 9 5.2 Muon Identification........................ 10 1

22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 5.2.1 Loose Muon Identification................ 10 5.2.2 Tight Muon Identification................ 10 5.3 Hadronic Tau Identification................... 10 6 Event Selection 10 6.1 Z selection............................. 10 6.1.1 Z µµ.......................... 10 6.1.2 Z ee........................... 11 6.2 Signal lepton selection...................... 11 6.2.1 Signal Electron Selection................. 11 6.2.2 Signal Muon Selection.................. 11 6.2.3 Signal Hadronic Tau Selection.............. 12 6.3 Higgs candidate selection..................... 12 6.3.1 h eµ.......................... 12 6.3.2 h eτ h.......................... 12 6.3.3 h µτ.......................... 13 6.3.4 h τ h τ h.......................... 13 6.4 Additional Requirements..................... 14 7 Background Estimation 15 7.1 Irreducible Background Estimation............... 15 7.2 Reducible Background Estimation................ 15 7.2.1 Electron and Muon Fake Rate Measurement...... 16 7.2.2 Hadronic Tau Fake Rate Measurement......... 16 7.2.3 Background Estimation Using Fake Rates....... 18 8 Reconstruction of A Mass 18 9 Systematic Uncertainties 19 10 Results 19 11 Kinematic Distributions 21 49 50 51 1 Introduction In the Yang-Mills theory of weak physics, all the gauge bosons are required to be massless. However, it was clear experimentally that the masses of the 2

52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 weak force mediators (W +/, Z) are on the order of 80 GeV. The process of electroweak symmetry breaking was proposed in order to provide mass to the weak force mediators. This theory introduces a new complex doublet. Through the process of spontaneous symmetry breaking, the degrees of freedom in this complex doublet are reduced from four to one. This remaining degree of freedom corresponds to a new scalar field with a non-zero vacuum expectation value, the Higgs field. The Higgs field implied the existence of new particle, the Higgs boson, which is an excitation of the Higgs field. However, the mass of the Higgs boson is an unspecified parameter in the theory. Since its proposal, experimental physicists have been searching for evidence of a Higgs boson. Until very recently, however, the particle remained elusive. The Large Hadron Collider (LHC) was primarily built to facilitate the discovery of the Higgs boson. The LHC, twenty-seven kilometers in circumference, collides proton beams traveling at nearly the speed of light to simulate the energy scales typical of the earliest stages of the universe. Although the exact mass of the Higgs boson is not predicted by the Standard Model, physicists expected that the LHC would be powerful enough to provide conclusive evidence of the existence of the Higgs boson. On September 10, 2008, the LHC began collecting collision data. For several years experimental physicists crunched the data looking for signs of the Higgs particle. On July 2, 2012 the CMS and ATLAS experiments at the LHC confirmed the observation of a Standard Model Higgs-like boson with a mass of about 125 GeV [3]. The discovery was a momentous achievement in particle physics which required the efforts of thousands of scientists from across the globe. Although every experiment so far on the new particle has agreed with the Standard Model expectations for a Higgs boson, many further precision studies are planned for the next run of the LHC. These studies will search for some difference in the properties of this new particle from the expectations put forth by the Standard Model theory for a Standard Model Higgs boson. Any deviation would suggest the presence of new physics outside the Standard Model. Although the Standard Model has time and again been verified by experiment, it is clear that the Standard Model does not completely describe our physical universe. The presence of dark matter, for example, is a phenomenon with no Standard Model explanation. Another significant issue with the Standard Model is the mass hierarchy problem, which requires the fine-tuning of some Standard Model parameters to 32 decimal places in order to give the Higgs boson the correct mass. 3

90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 Supersymmetry provides an elegant and simple solution to many of these issues. The supersymmetric model predicts the existence of a bosonic superpartner for each Standard Model fermion and a fermionic superpartner for each Standard Model boson. This new symmetry of nature effectively doubles the number of particles in the universe. If this symmetry were unbroken, each Standard Model particle would have a superpartner with the same mass. If this were the case, however, supersymmetric particles would have been discovered long ago. It is therefore necessary that supersymmetry is a broken symmetry. Unfortunately, this makes the masses of the supersymmetric particles unspecified. There are compelling arguments, however, that suggest that the lightest supersymmetric particles should have masses on the 1 TeV scale [7]. For this reason, many physicists expect to experimentally observe signs of supersymmetric particles during LHC Run 2. The most basic extension of the Standard Model which includes supersymmetry is the Minimal Supersymmetric Standard Model (MSSM). MSSM requires the existence of two complex scalar Higgs fields. This leads to two neutral CP-even Higgs particles, h 0 and H 0, a CP-odd Higgs particle A, and two charged and CP-even Higgs particles, H + and H. The masses of these five Higgs particles can be specified by two independent parameters, the mass of the A and tanβ, defined as: tanβ = v 1 /v 2 (1) where v 1 and v 2 are the vacuum expectation values for the neutral component of the Higgs field which couples to up and down quarks, respectively [7]. Figure 1 shows the predicted branching fractions of the A for various possible A masses with tanβ = 3 [9]. The decay of the A into a Z boson and h 0 has a relatively high branching fraction in the mass range between 260 GeV and 340 GeV. For an A mass less than 260 GeV, the A predominately decays to neutralinos. For an A mass greater than 340 GeV, A decay is dominated by t t. Therefore, it is convenient to probe this mass range using the subdominant A Zh mode. This analysis searches for an A which decays to a Z and an h 0, where the Z boson decays to e + e or µ + µ and the Higgs boson decays to τ + τ. The methodology of this analysis is very similar to the search for Standard Model Higgs associated production with a Z boson using tau final states [4]. Many of the techniques used in this analysis were also used in this Standard Model Higgs search. This thesis outlines all these methods, but focuses on 4

Figure 1: Predicted branching ratios of the A as a function of its mass with tanβ = 3. This analysis focuses on the mass range between 260 GeV and 340 GeV. 125 126 127 128 129 130 131 132 133 134 135 136 137 138 the methodology unique to this analysis.. 2 The CMS Detector The Compact Muon Solenoid detector consists of a 3.8 Tesla superconducting solenoid 6 meters in diameter. Within the superconducting solenoid are the brass scintillator hadronic calorimeter (HCAL), the electromagnetic calorimeter (ECAL) made of lead tungstate crystal, and a silicon pixel and strip tracker. Gas ionization chambers embedded in the steel flux return yoke outside the solenoid detect muons. The Level 1 Trigger uses the output from these detectors to select only the most interesting events. This selection process reduces the event rate from about 40 MHz to about 100 khz. The High Level Trigger (HLT) further reduces the event rate to about 300 Hz. These remaining events are stored for analysis purposes. For a complete description of the CMS apparatus refer to Ref[5]. 5

139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 3 Data and Monte Carlo Samples This analysis uses the DoubleElectron and DoubleMuParked primary dataset, collected by CMS in 2012 and corresponding to s = 8 TeV. These datasets require two loosely selected electrons or muons. The analysis also uses Monte Carlo simulated data to model the expected contributions from diboson production and Standard Model Higgs associated Z production. The Monte Carlo simulated datasets used in this analysis, along with the respective cross sections, are listed in Table 1. The expected signal contribution is modeled using FastSim software which incorporates Pythia8. FastSim signal samples are generated for m A between 260 GeV and 340 GeV in intervals of 10 GeV. Figure 3 compares some major kinematic variables for the FastSim sample versus a FullSim Pythia6 sample corresponding to m A = 300 GeV. These comparison plots show very good agreement, which demonstrates that the FastSim samples provide a useful initial simulation. FullSim MadGraph samples which correctly incorporate the Z polarization are currently being processed. Since these are not yet available, this analysis uses FastSim samples to model the signal. Process MC Generator σ (pb) Sample Name SM Zh llτ τ PYTHIA 0.0179 WH ZH TTH HToTauTau M-125 ZZ 4l PYTHIA 0.182 ZZTo4L TuneZ2star Table 1: Monte Carlo simulated datasets used for the irreducible background processes in this analysis. 156 157 158 159 160 4 Trigger Selection Triggers are selected which maximize the signal selection efficiency. Monte Carlo events are scaled to data with trigger efficiency scale factors calculated via the standard tag and probe method. Table 2 outlines the trigger paths used in this analysis 6

7

Figure 2: Comparisons of important kinematic variables for the FastSim versus the FullSim signal samples for m A = 300 GeV. The samples agree very well in all variables. The FastSim samples will be used in the remainder of this analysis because more mass points are currently available. HLT path µµ channels Mu17 Mu8 Mu17 TkMu8 ee channels Ele17 CaloIdT CaloIsoVL TrkIdVL TrkIsoVL Ele8 CaloIdT CaloIsoVL TrkIdVL TrkIsoVL Table 2: 2012 HLT paths used for this analysis. 8

161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 5 Particle Identification Electrons, muons, and hadronic taus are selected using the working points defined in the Standard Model h ττ search [4]. These selections have been optimized to obtain the best possible sensitivity for a Standard Model Higgs boson. Since the A decays to a Z boson and a Standard Model-like Higgs boson h 0, these selections should provide good sensitivity for this MSSM search as well. 5.1 Electron Selection Electron identification uses a Boosted Decision Tree (BDT) discriminator. The BDT is trained with simulated data and takes many kinematic variables as input. For a more complete description of these techniques refer to the relevant references [8]. 5.1.1 Loose Electron Identification In addition to the selection criteria used in the inclusive Standard Model h τ τ search, this analysis uses a looser electron working point. This working point was also used in the Standard Model Zh analysis. It is defined in Table 3. BDT Discriminator Value (>) η < 0.8 0.8 η < 1.479 1.479 η p T > 10 GeV (ZH) 0.50 0.12 0.60 Table 3: Thresholds for the BDT discriminator to identify loose electrons. For an identified electron the discriminator value has to fall above the indicated threshold. 178 179 180 181 5.1.2 Tight Electron Identification In some cases electrons are more selectively identified. These tighter selections, which correspond to the standard electron identification used in the Standard Model Higgs ττ search, are outlined in Table 4 9

BDT Discriminator Value (>) η < 0.8 0.8 η < 1.479 1.479 η p T 20 GeV (ZH) 0.925 0.915 0.965 p T > 20 GeV (ZH) 0.905 0.955 0.975 Table 4: Thresholds for the BDT discriminator to identify tight electrons. For an identified electron the discriminator value has to fall above the indicated threshold. 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 5.2 Muon Identification Muons are selected using the particle-flow algorithm detailed in [1]. 5.2.1 Loose Muon Identification In addition to passing the loose particle-flow selections, loose muons must also be identified as both Global and Tracker via the algorithms outlined in [2]. 5.2.2 Tight Muon Identification Tight muons must pass tight particle-flow selections, as recommended by the 2012 Muon POG [1]. 5.3 Hadronic Tau Identification Hadronic taus are identified using the Hadron Plus Strips (HPS) algorithm [6]. 6 Event Selection 6.1 Z selection Z boson candidate events are selected as follows: 6.1.1 Z µµ Two muons with the following properties: 10

199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 p T > 20 (10) GeV and η < 2.4 and Z < 0.1 Loose Muon Identification (Section 5.2.1 Relative PF β Isolation < 0.3 Opposite Sign µµ invariant mass between 60 GeV and 120 GeV 6.1.2 Z ee Two electrons with the following properties: p T > 20 (10) GeV and η < 2.5 and Z < 0.1 Loose Electron Identification 5.1.1 Relative PF β Isolation < 0.3 Opposite Sign ee invariant mass between 60 GeV and 120 GeV 6.2 Signal lepton selection The following selections are used to loosely select leptons as decay candidates from a tau which is a daughter of a Higgs boson. These selections are used for all channels. Tighter isolation and identification requirements, outlined in the next section, vary by channel. 6.2.1 Signal Electron Selection p T > 10GeV and η < 2.5 6.2.2 Signal Muon Selection p T > 10GeV and η < 2.4 11

220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 6.2.3 Signal Hadronic Tau Selection Decay Mode Finding (Section 5.3) p T > 15GeV and η < 2.3 6.3 Higgs candidate selection Signal lepton selections refer to the selections detailed in the previous section, while loose and tight lepton identification refers to the selections outlined in Section 4. 6.3.1 h eµ In this fully leptonic channel, each tau decays to a lepton and two neutrinos. This channel has the lowest predicted branching fraction. Signal muon and electron selection Loose electron and muon identification Electron and Muon Relative PF β Isolation < 0.3 Electron missing hits 1 Opposite sign Scalar sum of the p T of the electron and muon > 25 GeV 236 237 238 239 240 241 242 243 6.3.2 h eτ h In this channel, one tau decays hadronically and one tau decays to an electron and two neutrinos. Tighter selections are used in this channel in order to reduce the relatively high expected background contributions from WZ and Z+jets. Signal electron and tau selection Electron Relative PF β Isolation < 0.15 Tight electron identification 12

244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 No electron missing hits Tau Loose Combined 3 Hits Isolation Tau Anti-Muon Discriminator ( AntiMuonLoose2 ) Tau Anti-Electron Discriminator ( AntiElectronMVA3Tight ) Opposite Sign Scalar sum of the p T of the electron and tau > 30 GeV 6.3.3 h µτ In this channel, one tau decays hadronically and one tau decays to a muon and two neutrinos. The presence of a muon makes this channel relatively clean. Signal muon and tau selection Muon Relative PF β Isolation < 0.25 Loose Muon Identification Tau Loose Combined 3 Hits Isolation Tau Anti-Muon Discriminator ( AntiMuonTight2 ) Tau Anti-Electron Discriminator ( AntiElectronLoose2 ) Opposite Sign Scalar sum of the p T of the muon and tau > 45 GeV 262 263 264 265 266 267 6.3.4 h τ h τ h In this channel, both taus decay hadronically. This channel has a relatively high background contribution from Z+jets events, where two hadronic jets in the event fake hadronic taus. Events in this channel must contain two τ s which pass the following selections: Signal tau selection 13

268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 Medium Combined 3 Hits Isolation Anti-Muon Discriminator ( AntiMuonLoose2 ) Anti-Electron Discriminator ( AntiElectronLoose ) Opposite Sign Scalar sum of the p T of the two tau s > 70 GeV 6.4 Additional Requirements In addition to the selections outlined above for various channels, all events are required to pass the following selections: No extra muon which passes the following: Loose Muon Identification (Section 5.2.1) Signal Muon Selection Relative PF β Isolation < 0.3 R > 0.4 with respect to other leptons in the event No extra electron which passes the following: Loose Electron Identification (Section 5.1.1) Signal Electron Selection Relative PF β Isolation < 0.3 R > 0.4 with respect to other leptons in the event No b-jet which passes the following: p T > 20 and η < 2.4 bdiscriminator(ćombinedsecondaryvertexbjettags ) > 0.679 R > 0.4 with respect to other leptons in the event None of the four leptons within R = 0.1 of one another All leptons are within 0.1cm of the primary vertex. 14

Figure 3: Standard Model Higgs associated production 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 7 Background Estimation 7.1 Irreducible Background Estimation The predominant source of irreducible background is Standard Model ZZ production, where one Z decays to τ + τ the other Z decays to e + e or µ + µ. The process yields exactly the same final states as the expected signal. Another significant source of irreducible background in this analysis is Standard Model Higgs associated production with a Z boson (Figure 3). In this process, an off-shell Z radiates a Standard Model Higgs boson. When the Z decays to light leptons and the Higgs decays to τ + τ, the final states are indistinguishable from signal events. The expected contribution from both these processes is estimated using Monte Carlo simulated data generated with PYTHIA, as discussed in Section 3. 7.2 Reducible Background Estimation The primary source of reducible background is Z + jets. Another significant source is Standard Model WZ production. In both processes, one or more jets are misidentified as leptons. The contribution from these processes to the final selected events is estimated using a data-driven fake rate scheme. The probability of a jet faking a lepton, the fake rate, is measured in the same sign region. In this region, events are required to pass all the final state selections, except that the reconstructed tau candidates are required to have the same sign rather than opposite sign. This effectively eliminates any possible signal while maintaining roughly the same proportion of background events. 15

316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 7.2.1 Electron and Muon Fake Rate Measurement Electron (muon) fake rates are measured using eτ h (µτ h ) final states. Electrons (muons) are selected as outlined in Section 5 for the eτ h (µτ h ) final states, with the following exceptions: no isolation requirement no identification (Section 5.1.1) signal tau selected for p T > 5 GeV (instead of 15 GeV) no cut on the scalar p T sum electron (muon) and tau have the same sign electron (muon) MtToMET < 30 GeV to suppress real leptons from WZ and ZZ These looser selections allow for better statistics. Events that pass these selections define the denominator region. Electrons (muons) that also pass the loose identification and isolation requirements are included in the numerator region. The fake rate is calculated as the ratio of the number of events in the numerator region to the number of events in the denominator region. The fake rate is measured for ranging values of the closest jet p T, then fitted with a falling exponential. The best-fit exponential function is used to estimate the fakerate, F (jetp T ) for a given data event. These fits are shown in Figure 4. 7.2.2 Hadronic Tau Fake Rate Measurement The hadronic tau fake rate is measured from the τ h τ h channels. The selections are the same as those outlined in Section 5, with the exception that the cut on the scalar p T sum has been reduced to 50 GeV. The fake rate is calculated as the ratio of the number of events that pass all selections to the number of events that pass all selections other than isolation. As is done for electrons and muons, this fake rate is measured for various bins of closest jet p T, then fitted with a falling exponential. Figure 5 shows the calculated fake rate as a function of closest jet p T. 16

Figure 4: Electron (left) and muon (right) fake rates parameterized by closest jet p T. Figure 5: The hadronic tau fake rate parameterized by closest jet p T and fitted with a falling exponential for combined medium 3 hits tau isolation. 17

345 346 347 348 349 350 7.2.3 Background Estimation Using Fake Rates Data events are split into the three following categories and assigned the following weights: Category 0. Events that fail isolation or identification requirements on both tau candidate legs. This category is dominated by Z+jets. These events are assigned the weight F (τ 1 )F (τ 2 )/(1 F (τ 1 ))(1 F (τ 2 )) (2) 351 352 353 354 355 356 357 358 Category 1. Events that fail isolation or identification requirements on the first tau (the higher p T tau in ττ events, the electron in eµ events, and the electron (muon) in eτ(µτ) events) but pass for the second tau. This category includes Z+jets and WZ. These events are assigned the weight F (τ 1 )/(1 F (τ 1 )) (3) Category 2. Events that pass selections for the first tau but fail isolation or identification for the second tau. This category includes Z+jets and WZ. The events are assigned the weight F (τ 2 )/(1 F (τ 2 )) (4) 359 360 361 362 363 364 365 366 367 368 369 370 371 The reducible background is estimated as the weighted combination of categories 1 and 2 with category 0 subtracted. This combination of categories avoids double-counting of Z+jets events. Table 5 shows the contributions from each category split by channel. 8 Reconstruction of A Mass The Standard Model h ττ search used a special algorithm (SVFit) to reconstruct the τ τ invariant mass. For a complete description of this algorithm, refer to [4]. Unlike the Standard Model search, this analysis uses Markov Chain integration to extract the p T, η, and φ of the SV-fitted ττ system in addition to its invariant mass. The reconstructed A candidate mass is calculated as the invariant mass of the four-vector sum of the Z candidate and the ττ system. Figure 6 compares the area normalized shape of the visible mass and the A candidate mass for the signal and background samples. 18

channel PPPP 0 1 2 1+2-0 MMTT 5.0 1.176 (23065.0) 0.932 (263.0) 3.265 (172.0) 3.01 ± 0.57 MMET 6.0 8.524 (14983.0) 13.045 (420.0) 1.612 (117.0) 6.14 ± 0.93 MMMT 14.0 0.51 (3810.0) 1.104 (66.0) 1.667 (151.0) 2.26 ± 0.54 MMEM 10.0 0.394 (933.0) 0.779 (33.0) 1.163 (43.0) 1.55 ± 0.57 EETT 13.0 1.243 (24861.0) 1.122 (310.0) 4.037 (195.0) 3.92 ± 0.6 EEMT 11.0 0.536 (4337.0) 1.997 (91.0) 1.803 (164.0) 3.26 ± 0.6 EEET 17.0 3.093 (15377.0) 4.163 (387.0) 5.05 (341.0) 6.12 ± 0.71 EEEM 7.0 0.515 (999.0) 1.092 (46.0) 1.452 (70.0) 2.03 ± 0.59 Table 5: Reducible background counts in each channel and category. These contributions are estimated using the data-driven fake rate method detailed above. The right-most column represents the estimated reducible background contribution in each channel. 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 The reconstructed A mass greatly improves the shape difference between the signal (A Zh) and backgrounds, allowing for much better sensitivity in this analysis. 9 Systematic Uncertainties Systematic uncertainties are very similar to those used in the SM Zh search [4]. Table 9 lists the systematic uncertainties used in this analysis and their respective values. Table 7 shows the output of a maximum likelihood fit of these systematic uncertainties to the data. None of the nuisance parameters vary by more than half a standard deviation, which provides strong evidence that these uncertainties sufficiently model the data. 10 Results Figures 7 through 10 show the reconstructed A candidate mass by channel. Figure 11 shows the A candidate mass combined over all channels. This mass shape is used in the asymptotic limit calculations. Figures 12 through 15 show the 95% confidence level asymptotic upper limits on the gg A cross section times branching fraction in the mass range 260 GeV m A 340 GeV by channel. Figure 16 shows upper limits com- 19

Figure 6: A comparison of the area normalized mass shapes for the visible mass vs. the A candidate mass discussed in Section 8. The blue histogram corresponds to the signal (A Zh) FastSim MC for ma = 300 GeV. The red histogram corresponds to the ZZ irreducible background, which is extracted from MC simulated data. The green histogram corresponds to the reducible background, which is calculated using the data-driven fake rate method outlined in Section 7.2. The reconstructed mass greatly improves the shape separation between the signal and backgrounds, allowing for significantly improved sensitivity. 20

Systematic uncertainties common to all channels. Source Uncertainty Luminosity measurement 2.6% Muon trigger efficiency 1% Muon ID/Iso/ES 2% Electron trigger efficiency 1% Electron ID/Iso/ES 2% Tau ID/Iso 6%(12%) Tau ES 3%(6%) Btag 1% PDF for q q ZZ 5% QCD scale for q q ZZ 6% QCD scale for VHs 3.1% Reducible background estimate 10-30% Table 6: Systematic uncertainties. The uncertainties on e and µ reconstruction and identification (ID), are isolation (Iso) are combined; for τ h, the energy-scale uncertainty is reported separately. 389 390 391 392 393 394 395 396 397 398 399 400 401 bined over all channels. No excess is observed, although further sensitivity is required to probe the interesting MSSM parameter space. 11 Kinematic Distributions The following plots compare kinematic distributions from data with the expected background contributions, to illustrate the quality of this analysis. The Standard Model ZZ and Standard Model Zh expected contributions, estimated using Monte Carlo simulated data, are drawn in red and dashed black, respectively. These samples are weighted to match the integrated luminosity in data. The expected reducible background contribution, calculated using the data-driven fake rate method detailed in section 7.2, is drawn in green. The data is overlayed with the sum of the background shapes. These plots have NOT been fitted, yet they show good agreement between the observed data and Monte Carlo simulation. 21

b-only fit s + b fit name x/σ in, σ out /σ in x/σ in, σ out /σ in ρ(θ, µ) CMS eff e 8TeV +0.53, 0.96 +0.53, 0.96-0.00 CMS eff m 8TeV -0.15, 0.97-0.15, 0.97-0.00 CMS eff t llet 8TeV -0.10, 0.97-0.10, 0.97-0.00 CMS eff t llmt 8TeV +0.35, 0.97 +0.35, 0.97-0.00 CMS eff t lltt 8TeV -0.11, 0.90-0.11, 0.90-0.00 CMS fake b 8TeV +0.06, 0.99 +0.05, 0.99-0.00 CMS trigger e 8TeV +0.13, 0.99 +0.13, 0.99-0.00 CMS trigger m 8TeV -0.08, 0.99-0.08, 0.99-0.00 CMS vhtt zznorm em 8TeV +0.12, 0.99 +0.12, 0.99-0.00 CMS vhtt zznorm et 8TeV +0.00, 0.99 +0.00, 0.99-0.00 CMS vhtt zznorm mt 8TeV +0.21, 0.99 +0.20, 0.99-0.00 CMS vhtt zznorm tt 8TeV -0.03, 0.99-0.03, 0.99-0.00 CMS zh2l2tau ZjetBkg emu extrap 8TeV -0.07, 0.94-0.08, 0.94-0.00 CMS zh2l2tau ZjetBkg lt extrap 8TeV +0.01, 0.85 +0.01, 0.85-0.00 CMS zh2l2tau ZjetBkg tt extrap 8TeV -0.05, 0.97-0.05, 0.97-0.00 QCDscale VH +0.00, 0.99-0.00, 0.99-0.00 QCDscale VV +0.36, 0.98 +0.36, 0.98-0.00 lumi 8TeV +0.14, 0.98 +0.14, 0.98-0.00 pdf qqbar +0.30, 0.98 +0.30, 0.98-0.00 Table 7: Nuisance parameter variation after performing a maximum likelihood fit on data. The variation in the nuisance parameters is relatively small. This provides strong evidence that the systematic uncertainties presented above sufficiently model the data. 22

Figure 7: Comparison of SV-fit reconstructed A candidate mass expected contributions with data in the eµ channels. Z ee (left) and Z µµ (right). Figure 8: Comparison of SV-fit reconstructed A candidate mass expected contributions with data in the eτ channels. Z ee (left) and Z µµ (right). 23

Figure 9: Comparison of SV-fit reconstructed A candidate mass expected contributions with data in the µτ channels. Z ee (left) and Z µµ (right). Figure 10: Comparison of SV-fit reconstructed A candidate mass expected contributions with data in the ττ channels. Z ee (left) and Z µµ (right). 24

Figure 11: Comparison of SV-fit reconstructed A candidate mass expected contributions with data combined over all channels. 25

Figure 12: 95% confidence level upper limits on the A Zh llττ cross section times branching fraction for the eµ channels. Z ee (left) and Z µµ (right). Figure 13: 95% confidence level upper limits on the A Zh llττ cross section times branching fraction for the eτ h channels. Z ee (left) and Z µµ (right). 26

Figure 14: 95% confidence level upper limits on the A Zh llττ cross section times branching fraction for the µτ h channels. Z ee (left) and Z µµ (right). Figure 15: 95% confidence level upper limits on the A Zh llττ cross section times branching fraction for the τ h τ h channels. Z ee (left) and Z µµ (right). 27

Figure 16: 95% confidence level upper limits on the gga cross section times A Zh llττ branching fraction compared with the expected contributions from some interesting MSSM models. 28

Figure 17: Comparison of expected background contribution with data for τ 1 kinematics combined over all eight channels Figure 18: Comparison of expected background contribution with data for τ 2 kinematics combined over all eight channels 29

Figure 19: Comparison of expected background contribution with data for reconstructed Z kinematics, combined over all eight channels 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 References [1] Particle-Flow Event Reconstruction in CMS and Performance for Jets, Taus, and MET. Technical Report CMS-PAS-PFT-09-001, CERN, 2009. Geneva, Apr 2009. [2] Performance of muon identification in pp collisions at s**0.5 = 7 TeV. Technical Report CMS-PAS-MUO-10-002, CERN, 2010. Geneva, 2010. [3] Serguei Chatrchyan et al. Observation of a new boson at a mass of 125 GeV with the CMS experiment at the LHC. Phys.Lett., B716:30 61, 2012. [4] Serguei Chatrchyan et al. Evidence for the 125 GeV Higgs boson decaying to a pair of τ leptons. 2014. [5] Chatrchyan, Serguei, and others. The CMS experiment at the CERN LHC. Journal of Instrumentation, 3(08):S08004, 2008. [6] CMS Collaboration. Performance of -lepton reconstruction and identification in cms. Journal of Instrumentation, 7(01):P01001, 2012. [7] Howard E. Haber. Introductory low-energy supersymmetry. 1993. 30

Figure 20: Comparison of expected background contribution with data for reconstructed τ τ events, combined over all eight channels 31

418 419 420 421 [8] Andreas Hocker, J. Stelzer, F. Tegenfeldt, H. Voss, K. Voss, et al. TMVA - Toolkit for Multivariate Data Analysis. PoS, ACAT:040, 2007. [9] Ian M. Lewis. Closing the Wedge with 300 fb 1 and 3000 fb 1 at the LHC: A Snowmass White Paper. 2013. 32