Searching for the Higgs at the Tevatron 5 th May 2009 Particle Physics Seminar, Oxford
Outline Introduction The challenges and analysis strategies Low mass SM Higgs searches High mass SM Higgs searches Combined Tevatron result Conclusions 2
The Standard Model of Particle Physics 12 particles and antiparticles: quarks and leptons Carriers of forces: vector bosons 3
What is the origin of mass? A very basic question is why particles have masses... they interact with the Higgs field 4
The Higgs Boson In the Standard Model, the Higgs field is a complex scalar field, V(ϕ) Through electroweak symmetry breaking, the gauge bosons (W, Z) acquire mass A single Higgs boson with spin 0 appears. The only free parameter is its mass 5
The Higgs Boson In the Standard Model, the Higgs field is a complex scalar field, V(ϕ) Through electroweak symmetry breaking, the gauge bosons (W, Z) acquire mass A single Higgs boson with spin 0 appears. The only free parameter is its mass The Higgs particle has been searched for for decades, but not yet found 6
LEP s Legacy e - e + s - mz = 206.6-91.2 = 115.4 GeV Search for the Standard Model Higgs boson at LEP Phys. Lett. B565 (2003) 61. 7
LEP s Legacy Construct MC toy experiments for background only or signal plus background hypotheses - Incorporate systematic uncertainties through Gaussian smearing Evaluate the statistical significance with a Poisson log likelihood ratio (obtain probability densities) 1-CLB : probability for a signal like fluctuation of the background Excess around 115 slightly more s+b like mh 114.4 GeV @ 95% CL 1-CLB CLSB 8
Stalking the Higgs The SM relates mh, mt, mw via radiative corrections: New precision measurements on the top and W mass from the Tevatron Indirect constraints on the Higgs boson mass from global EW fits: mh = 90 ± 36 27 GeV mh < 191 GeV @95%CL A light Higgs boson might be around the corner (if the SM is correct)! 9
Tevatron collider in Run II 10
Tevatron collider in Run II Run I 1987-1995 Run II ongoing since 2001 Delivers a data set equal to Run I (~100 pb -1 ) every 2 weeks Current high energy frontier! 11
CDF and DØ experiments in Run II Two General-Purpose Detectors: Precision tracking Hermetic calorimeter Muon system Both detectors are highly upgraded in Run II Well understood, stable operation and efficient data taking 12
Data set Up to 4.2 fb -1 of data analysed (after data quality requirements) Expected to ~double this data set by the end of 2010 Similar for CDF 13
Tevatron performance More data with higher instantaneous luminosities 14
Challenges with high luminosity A zero bias event @ 60E30 cm 2 s -1... and @ 240E30 cm 2 s -1 Average number of interactions: LHC: initial low lumi run (L=2000E30 cm 2 s -1 ): <N>=3.5 TeV: (L=300E30 cm 2 s -1 ): <N>=10 15
Challenges with high luminosity A zero bias event @ 60E30 cm 2 s -1... and @ 240E30 cm 2 s -1 Average number of interactions: LHC: initial low lumi run (L=2000E30 cm 2 s -1 ): <N>=3.5 TeV: (L=300E30 cm 2 s -1 ): <N>=10 To cope with high luminosities: New techniques to improve calibration Improve / redesign algorithms for electron, photon, jet, tau and missing transverse momentum reconstruction 16
Higgs production at the Tevatron Events only one in ~10 12 events will be a Higgs boson 17
Higgs production at the Tevatron Events Discovery March 09 only one in ~10 12 events will be a Higgs boson 17
Higgs production at the Tevatron Events only one in ~10 12 events will be a Higgs boson Comparing σtot/σhiggs with a typical density of a haystack r = 620m 18
Higgs production at the Tevatron Events (pb) Higgs production cross sections are small: 0.1-1 pb depending on mh only one in ~10 12 events will be a Higgs boson Dominant production mode is gluon fusion 19
Higgs decays 135 GeV 20
Search strategy at the Tevatron 135 GeV Investigate different production mechanisms and a large number of final states 21
Search strategy at the Tevatron 135 GeV mh > 135 GeV: gg H production with decay to WW 22
Search strategy at the Tevatron 135 GeV mh < 135 GeV: Associated production WH and ZH with H bb decay mh > 135 GeV: gg H production with decay to WW 23
Searches for a low mass Higgs mh < 135 GeV: Associated production WH and ZH with H bb decay 24
Main low mass search channels MET+l+bb: WH lνbb Largest VH production cross section More backgrounds than ZH llbb ll+bb: ZH llbb Less backgrounds Fully constrained Smallest Higgs signal MET+bb: ZH ννbb 3x more signal than ZH llbb (+WH lνbb when lepton missing) Large backgrounds which are difficult to handle 25
Signal and backgrounds Experimental signature: W*/Z* Missing transverse energy and/or isolated leptons Two high pt jets, acoplanar, b-tagged Main backgrounds: - Physics (from MCs): W/Z+jets, diboson, tt and single top - Instrumental (from Data): Multijet events with mismeasured missing E T or jets faking leptons Constrain and test background modelling in sideband regions 26
Searches for low mass Higgs A three step approach. Example: ZH ννbb large MET 2 high pt b-jets First step: select events consistent with W/Z and 2 jets Z+jets Δϕ Multijet W+jets Signal x500 27
Searches for low mass Higgs Second step: b-tagging Exploit B meson lifetime, mass, fragmentation and decay modes to separate b from light-quark jets - secondary vertex - tracks impact parameter - vertex track multiplicity - vertex mass - soft leptons Similar for CDF Use neural networks for optimal combination of tagging information 28
Searches for low mass Higgs After b-tagging Backgrounds dominated by: W/Z+bb and top Most discriminating quantity: dijet invariant mass Z+bb W+bb Signal x10 top 29
Dijet mass resolution Improving the dijet mass resolution crucial for low mass Higgs searches at the Tevatron (limits improve with the same relative rate) Use information from the tracker to correct calorimeter measurement (e.g. 10% improvement for CDF s ZH ννbb search) Correct for muons in jets and for semileptonic decays (e.g. 10% improvement for DØ s ZH ννbb + other low mass channels) Calorimeter shape, energy density and pre-shower information Correlations among all these 30
Searches for low mass Higgs Third step: Optimise separation power with multivariate discrimination Most common techniques: Neural Network, Decision Tree and Matrix Element - Exploit information from several final state variables and correlations among them Decision Tree discriminant 31
ZH ννbb Main challenge: modelling and rejection of large multijet background Understand modelling of main backgrounds in control regions: - Multijet enhanced: MET aligned with jets - EW, top enhanced: require isolated lepton Multijet CR EW, top CR 50% of the signal from WH (l)νbb Signal region Once control regions well understood, open the box perform search in signal region 32
ZH ννbb Reject multijet events with cuts on topological variables Extract final cross section limit from Decision Tree output mh=115 GeV Exp/Obs: 7.5/8.4 x σsm Reject multijet events with NN Use 2 nd NN to separate remainder backgrounds and extract cross section limits mh=115 GeV Exp/Obs: 5.6/6.9 x σsm 33
ZH llbb 2 isolated leptons 2 high pt b-jets Less backgrounds but also smallest Higgs signal Fully reconstructed, no real missing ET 34
ZH llbb 2 isolated leptons 2 high pt b-jets Less backgrounds but also smallest Higgs signal Fully reconstructed, no real missing ET Increase lepton acceptance with loose leptons from Z Use constraints to improve dijet mass resolution 34
ZH llbb Signal acceptance increased with loosening selection criteria for 2nd lepton - Selection of lepton pairs consistent with Z mass gives handle against uncontrolled increase of backgrounds CDF: events with electron + track pointing to gaps between CAL cells, events with muon + MIP without hit in muon chambers DØ: electrons in inter-cryostat region, events with muon + isolated track 35
ZH llbb Z, H: fully reconstructed, no intrinsic missing ET use constraints to improve di-jet mass resolution - CDF: train a NN to correct jets to parton level - DØ: perform a kinematic fit (constrain the lepton pair to the Z mass and transverse momentum conservation in ZH system) Gives up to ~10% improvement in the limits 36
ZH llbb Both, CDF and DØ, split data into several sub-samples with different S/B to increase sensitivity - Tight/loose lepton definitions - Number of tagged jets and b-tagging efficiency Final discrimination with multivariate techniques - DØ: Boosted decision trees - CDF: Two dimensional NN, discriminating against Z+jets and tt µµ - single b-tag mh=115 GeV CDF: Exp/Obs: 9.9/7.1 x σsm DØ: Exp/Obs: 8.0/9.1 x σsm 37
WH lνbb 2 jets 3 jets Most sensitive final state in low mass region 1 b-tag 2 b-tags DØ: 4 categories, number of jets and number of b-tags CDF: 3 categories according to b-tags 38
WH lνbb DØ: NN with kinematic variables and ME discriminant + 6 other input variables CDF: NN with kinematic variables + BDT with ME discriminant combined NN superdiscriminant 39
WH lνbb DØ: Sensitivity gain with NN ~15%, additional ~8% improvement with ME CDF: Additional ~15% improvement with NN superdiscriminant mh=115 GeV Exp/Obs: 6.4/6.7 x σsm mh=115 GeV Exp/Obs: 4.8/5.6 x σsm At high discriminant values S/B typically 1/10-1/20 for the most sensitive low mass channels 40
WH lνbb DØ: Sensitivity gain with NN ~15%, additional ~8% improvement with ME CDF: Additional ~15% improvement with NN superdiscriminant Main systematic uncertainties for low mass channels: - Signal (total 15%): cross section, b-tagging, ID efficiencies - Background (total 25-30%): normalisation of W/Z+jets heavy flavour samples, modelling of multijet and W/Z+jets background, b-tagging Systematic uncertainties reduce sensitivity of single channels by 20-30% (included in these limits) 41
Final states with taus H ττ: second largest BR in low mass region Two analyses: 1. Select events with ττjj final state. Sensitive to ZH (Z ττ, H bb), VH (V jj, H ττ) and vector boson/gluon fusion with H ττ 42
Final states with taus H ττ: second largest BR in low mass region Two analyses: 1. Select events with ττjj final state. Sensitive to ZH (Z ττ, H bb), VH (V jj, H ττ) and vector boson/gluon fusion with H ττ Combination of several NNs trained for different backgrounds and signals ττjj ττjj 43
Final states with taus H ττ: second largest BR in low mass region Two analyses: 1. Select events with ττjj final state. Sensitive to ZH (Z ττ, H bb), VH (V jj, H ττ) and vector boson/gluon fusion with H ττ Combination of several NNs trained for different backgrounds and signals ττjj ττjj τνbb 2. Events with the τνbb final state. Sensitive to WH τνbb Sensitivity of tau channels ~30 x σsm with 1-2 fb -1 44
Searches for a high mass Higgs mh > 135 GeV: gg H production with decay to WW 45
H+X l+l- + missing ET Dominant decay for mh>135 GeV: H W*W Clean environment can take advantage of gg H production Signal contribution also from W/Z+H, qqh production Consider all sources of opposite sign di-lepton + missing ET Backgrounds: Drell-Yan production, diboson, tt, W+jets/γ, multijet 46
H+X l + l - + missing ET Dominant background of Drell-Yan production reduced with cuts on missing ET and its significance Cut at 25 (15 for eµ) already applied Δϕ(l,l) Spin correlation gives main discrimination against irreducible background from nonresonant SM W pair production Signal WW 47
H+X l + l - + missing ET To increase sensitivity: DØ: Split the samples according to lepton flavour and combine result Neural Network with 11 kinematic and topological input variables Zoom CDF: Split samples into jet multiplicity and lepton ID criteria: different signal and background composition Veto events with tight b-tagged jet Neural Network with additional ME input for the 0-jet NN 48
Systematic uncertainties Main systematic uncertainties: - Signal (total 10%): cross section, lepton ID/ trigger - Background (total 13%): cross sections, jet lepton fake rate, jet ID/resolution/calibration Systematic uncertainties change rate and shape of the signal and background predictions SM signal expectation and data after background subtraction Constrained total systematic uncertainty Expected 165 GeV SM Higgs signal would be visible over background uncertainty 49
H+X l+l- + missing ET Exclusion limits per experiment: mh=165 GeV Exp/Obs: 1.7/1.3 x σsm mh=165 GeV Exp/Obs: 1.4/1.5 x σsm At high Neural Network values S/B close to 1/3 With additional luminosity and improvements (e.g. additional channels) expect single experiment exclusion around mh = 165 GeV in the near future 50
Intermediate mass region Important analyses in the intermediate (130-150 GeV) mass region: Search for associated production with WH WWW l ± νl ± ν+x decay - Remove physics backgrounds with like-sign lepton requirement and veto on additional leptons - CDF: further background rejection by requiring at least one jet ~7 x σsm ~18 x σsm Search for H γγ - Backgrounds: direct diphoton production (irreducible), dijet/photon+jet, DY - Small branching fraction (look for gluon fusion production) - Peaking signal compensates for low S/B 51
Limit setting Full combination of all channels from CDF and DØ for best sensitivity - Combining ~30 different channels per experiment - More than 50 different sources of systematic uncertainties are considered LEP: Tevatron/LHC: low background, small systematics high background, large systematics Degrading effect of systematic uncertainties larger at the Tevatron Similar methods for limit settings as used at LEP, but counteract the degrading effects from uncertainties via Profile Likelihood technique Likelihood now a function of nuisance parameters Background is constrained by maximising profile likelihood ( sideband fitting ) 52
Combined Tevatron limits 1σ (green) 2σ (yellow) stat.+syst. uncertainty on expected limit Expected limit Observed limit (data) 53
Combined Tevatron limits 1σ (green) 2σ (yellow) stat.+syst. uncertainty on expected limit Expected limit Observed limit (data) The Tevatron experiments reached sensitivity to the SM Higgs boson around the mass region of 160 170 GeV. First direct exclusion since LEP! At mh=115 GeV Exp/Obs: 2.4/2.5 x σsm Effective luminosity for current analyses in CDF+DØ combination at the low mass region: 2.55 fb-1. Reach of sensitivity will be expanded greatly 53
LHC discovery potential ATLAS as an example, similar conclusions from CMS 5σ mh < 130 GeV Main channels H γγ and H ττ Challenging, a few years of running may be needed mh > 130 GeV Main channels H WW and H ZZ Discovery with few fb -1 up to masses favored by the SM possible 54
LHC discovery potential It is likely that the Tevatron experiment will exclude a larger mass range around 160 GeV by the end of 2010 5σ mh < 130 GeV Main channels H γγ and H ττ Challenging, a few years of running may be needed mh > 130 GeV Main channels H WW and H ZZ Discovery with few fb -1 up to masses favored by the SM possible 54
LHC discovery potential It is likely that the Tevatron experiment will exclude a larger mass range around 160 GeV by the end of 2010 5σ mh < 130 GeV Main channels H γγ and H ττ Challenging, a few years of running may be needed In this region Higgs decays to bb at the Tevatron will provide complementary information to LHC results mh > 130 GeV Main channels H WW and H ZZ Discovery with few fb -1 up to masses favored by the SM possible 54
Conclusions The Tevatron experiments achieved sensitivity to the SM Higgs boson production cross section first direct exclusion since LEP! The Tevatron can deliver more than 9 fb -1 of data by the end of 2010 and the reach of sensitivity will be expanded greatly To be able to find evidence for the Higgs over the entire mass range preferred by EW fits will be challenging but well possible Very exiting times ahead at the Tevatron in the next 1-2 years! 55
Backup slides 56
Decision Trees Idea: recover events that fail criteria in cut-based analysis S/B S/B S/B 57
Matrix elements 58