David Futyan

1
The Discovery Potential of the Higgs Boson at CMS
in the Four Lepton Final State

• David Futyan
• UC Riverside

2
Overview

• Introduction
• LHC and CMS
• Motivation for Higgs boson searches
• Decay channels observable at the LHC
• Signal and background processes
• Cross-sections and branching ratios
• Event generation and simulation
• Online selection
• Offline reconstruction of electrons and muons
• Offline event selection
• Evaluation of background from data
• Significance with background systematics
• Potential for measurement of Higgs boson
  properties
• Mass, width, cross-section
• Experimental systematic uncertainties

3
The LHC (Large Hadron Collider)

• Proton-proton collider
• vs 14 TeV
• Luminosity 1034 cm-2s-1

• 17 miles in circumference
• Due to begin operation summer 2007

4
The CMS Detector (Compact Muon Solenoid)

• General purpose detector
• Over 2000 people from
• 160 institutes
• 12500 tonnes

5
CMS Detector Slice
6
CMS Under Construction
7
The Higgs Boson

• A key objective of the LHC is to elucidate the
  origin of mass.
• Higgs mechanism
• Provides an explanation for electroweak symmetry
  breaking in the Standard Model
• Gives rise to the massive Z and W vector bosons
  and the massless photon.
• gtLies at the core of the Standard Model -
  without the Higgs mechanism the SM is neither
  consistent nor complete.
• Provides mechanism through which gauge bosons and
  fermions acquire mass.
• Predicts the existence of one physical scalar,
  neutral Higgs boson.

8
Higgs Boson Mass Constraints

• Higgs boson mass not predicted by the theory -
  free parameter of the standard Model
• Must be determined experimentally.
• Current limits
• Combined lower limit from direct searches
• at LEP mH gt 114.4 GeV/c2 (95 CL).
• Higgs boson contributes to radiative
• corrections to electroweak observables.
• Consistency fits to electroweak
• precision measurements from LEP, SLC,
• Tevatron yield an indirect upper limit
• mH lt 207 GeV/c2 (95 CL).

9
The Higgs Boson Production at the LHC

• Dominant production mechanisms
• Gluon-gluon fusion contributes around 80 of the
  total
• Decay channels

10
LHC Search Channels for the Higgs Boson

• Decay channels of SM Higgs boson which yield
  highest sensitivity for discovery at the LHC
• H ? WW- ? 2l2?
• H ? ZZ() ? 4l
• H ? ??

l electron or muon
LEP Limit 114.4 GeV/c2
11
The H?ZZ?4l Channel

• Most sensitive channel for the discovery of the
  Higgs boson at the LHC for a wide range of
  masses.
• Exceptionally clean signature of 4 isolated high
  pT leptons, with relatively small backgrounds.
• For mHgt2mZ Golden channel, with 2 real Z
  bosons
• Mass of Higgs boson can be directly reconstructed
  from the invariant mass of the 4 leptons
• Direct measurement of mass
• Direct measurement of width for large mH (gt200
  GeV)

12
General Strategy

• Detailed analyses have been developed largely
  independently for each of the three final states
• 4e LLR (France), Split (Croatia), Rome/INFN
• 4???Florida, FNAL, Cambridge,
• 2e2? UC Riverside, Bari/INFN (Italy)
• Common to all channels (allows coherent
  combination of results)
• Event generation, detector simulation.
• Signal and background production and decay
  processes considered and their NLO cross-sections
  and branching ratios.
• Straight forward counting experiment approach -
  Cut based analyses
• Look for local event excess over expected
  background.
• Details of event selections developed differently
  in each of the 3 channels.
• Analyses are designed as if real data were being
  analyzed
• Full detailed simulation of CMS detector geometry
  and response.
• Simulation of LHC conditions in first years of
  running at L 21033cm-2s-1.
• Full treatment of systematic errors included in
  significance evaluation.
• Techniques developed to measure the size of the
  residual background from LHC data.

13
Production Cross-Section and Branching Ratio
Sum of gg fusion, WW fusion, ZZ fusion
BR(H?ZZ()?4l), including BR(t ?e,m)

• BR(H?ZZ()?4l) is the branching ratio to a final
  state containing only e and m, including t decay
  products.

14
Enhancement for 4e and 4? Final States

• For 4e and 4? final states, enhancement of signal
  cross-section due to constructive final state
  interference between like-sign electrons or muons
  originating from different Z() bosons

Calculated using CompHEP
15
Signal Monte Carlo Event Generation

• Signal samples generated with PYTHIA for 18 mass
  points between 115 and 600 GeV.
• Higgs production mechanisms simulated
• gg fusion, WW fusion, ZZ fusion.
• Z bosons forced to decay to e,m,t, with t forced
  to decay to e,??
• 10000 events generated per mass point, for each
  final state (4e, 4?, 2e2?)
• Events re-weighted to correspond to

where
16
Background Processes

• Reducible backgrounds
• qq/gg ??tt ??WW-bb ??4l X (PYTHIA)
• qq/gg ??(Z()/g) bb ? 4l X (CompHEP interfaced
  with PYTHIA)
• Irreducible non-resonant continuum background
• qq ???Z()/?)(Z()/g) ? 4l (PYTHIA)

Process ?LO(pb) NLO K-factor ?NLO(pb)
tt?WW-bb - - 840
ee-bb 115 2.4?0.3 276
????bb 116 2.4?0.3 279
?Z()/?)(Z()/g) 18.7 KNLO(m4l) 0.2 28.9

(Z()/g) bb
17
?Z()/?)(Z()/g) Background

• LO cross-section 18.07pb (from MCFM generator)

l
q
Z()/?
l-
q u,d,s,c or b
l
Z()/?
q
l-

• t-channel dominates
• 90 m4llt2mZ, 100 m4lgt2mZ
• s-channel simulated for 4? final state only
• MCFM generator used to calculate an K-factor to
  account for all NLO processes
• Function of 4 lepton invariant mass

18
?Z()/?)(Z()/g) Background

• Significant NNLO box diagram process
• not included in the simulation
• TOPREX generator used to obtain ratio
  ?(gg?ZZ?4l)/?(qq?ZZ?4l) 20
• Total NLO cross-section sLO (K(m4l) 0.2)
  29pb (for average K(m4l)1.35)
• All events re-weighted at analysis level using
  this m4l dependent K-factor.

19
Other Potential Backgrounds

• Zcc can also give 4 leptons in the final state
• Investigated with full detector simulation -
  found to be negligible
• Other potential sources of background
  investigated at generator level
• Wbb
• Wcc
• Single top
• bbbb
• bbcc
• cccc
• All found to be negligible

One or more fake leptons

All leptons non-isolated
20
Detector Simulation

• Generated Monte Carlo events for all generated
  samples are passed through a highly detailed
  simulation of the CMS detector, including
• Precise simulation of the complete detector
  geometry all material in the detector including
  cables, services etc.
• Detailed simulation of the 4T magnetic field.
• Full simulation of detector response for all
  detector components information used as input
  to the analysis fully simulates real LHC data.
• Generated events are mixed with pile-up events to
  simulate the LHC conditions at low luminosity
  (21033cm-2s-1)
• Several inelastic pp collisions per bunch
  crossing
• Corresponds to conditions during the initial
  phase of data taking.

21
Cross-Section Times Branching Ratio

• Generator level kinematic preselection includes
  the final state lepton flavor requirement (4e, 2?
  or 2e2?), plus generator level cuts
• Electrons pTgt5GeV, hlt2.5
• Muons pTgt3GeV, hlt2.4
• For 2e2? case

tt Zbb ZZ
s (fb) 840x103 555x103 28.9x103
s.BR.e (fb) 744 390 37.0
22
4-lepton Invariant Mass After Generator
Pre-selection
s-channel ZZ production
Same on linear scale
mH140 GeV signal
23
Online Selection

• LHC bunch crossing rate is 40MHz. Multiple
  events per bunch crossing
• CMS Trigger consists of a Level-1 trigger
  followed by a High Level Trigger. HLT is a
  software trigger involving full reconstruction of
  physics objects.
• Triggers chosen for H?ZZ?4l channels
• Single triggers were also considered for the 2e2?
  channel but were found not to benefit the final
  significance.

Channel Trigger
?? single electron double electron
?? Single muon double muon
???? double electron double muon
24
HLT Selection Efficiencies
2e2?
4e
tt 0.399 0.001 Zbb 0.661 0.001 ZZ 0.896
0.004

• For 4? channel, HLT efficiency is close to 100
  for all samples

25
Muon Reconstruction and Selection

• Muons are reconstructed with high efficiency with
  CMS
• Require ? and ?- reconstructed with pTgt7 in the
  barrel and pTgt13 in the endcaps.
• Require M(µµ-)gt12GeV for all permutations
  (excludes low mass resonances).
• These cuts have little effect on signal
  efficiency.

26
Electron Reconstruction

• Lowest pT electron in H?ZZ?4e events around
  10GeV
• Electrons radiate on average half their energy
  before reaching the ECAL due to
• Strong magnetic field (4 Tesla)
• 1 X0 of material in the inner tracker
• Energy is radiated as photons which may in turn
  convert to ee- before reaching the ECAL -
  significant spread of energy in ?.

27
Electron Reconstruction

• Sophisticated algorithms developed, motivated by
  the H?ZZ?4e analysis, in order to achieve good
  reconstruction efficiency for low pT electrons
• Use of Gaussian Sum Filter tracking - electron
  track is reconstructed right out to ECAL surface.
  Measure bremsstrahlung energy loss
• Categorization of electrons according to amount
  of radiated energy, ECAL cluster shape,
  cluster-track matching.
• Combine ECAL energy and tracker
• momentum measurements based on
• measurement uncertainties

28
Electron Selection

• Electron reconstruction has a significant
  background from fakes (e.g. p/p0 overlap from
  underlying event).
• Selection important to exclude potential
  backgrounds which can fake one or more electrons.
• 4e analysis Cut based selection
• Ecalo/pTrack lt 3.
• Track cluster matching ?? lt0.02 and ??lt0.1
• EHCAL/EECAL lt0.2
• pTgt 5 GeV
• Loose isolation ?pT/pT lt 0.5 (cone R0.2)
• 2e2? analysis
• Likelihood developed based on
• similar variables.
• Require likelihoodgt0.2.
• Select electron and positron with
• highest likelihoods.

29
Electron Reconstruction Efficiency (4e channel)
30
Offline Event Selection

• For the signal, and for the irreducible ZZ
  background, all 4 leptons are isolated and
  originate from the primary vertex.
• For the reducible tt and Zbb backgrounds, 2 of
  the leptons are associated with b-jets ?
  non-isolated and with displaced vertices.
• For all three channels, offline selection
  consists of two set do cuts
• Vertex/Impact parameter and Isolation cuts -
  reduce Zbb and tt only.
• Kinematic cuts lepton pT and lepton invariant
  mass cuts - reduce all backgrounds.
• The offline selection for the 2e2? channel is
  described on the following slides.

31
Vertex and Impact Parameter Cuts (2e2?)

• 3 variables chosen High background rejection
  for 95 signal efficiency. Largely uncorrelated
• (1) Transverse distance from mm- vertex to beam
  line lt 0.011 cm
• (2) 3D Distance between mm- and ee- vertices lt
  0.06 cm
• (3) Transverse impact parameter significance of
  lepton with highest IP significance lt 7

Combined Efficiency ()
Signal 89-91
tt 14.5 0.2
Zbb 13.0 0.1
32
Tracker Isolation (2e2?)

• Cut on SpT of all reconstructed tracks in the
  event which satisfy
• pTgt0.9 GeV
• At least 5 hits
• Within region defined as the sum of cones of size
  DRlt0.25 around each lepton, excluding veto cones
  of size DRgt0.015 around each lepton.
• Consistent with originating from the
  reconstructed primary vertex to within
  Dzlt0.2cm

33
Kinematic Distributions for Reconstructed Leptons

• Shown for events passing HLT and with ee-mm-
  reconstructed

34
Kinematic Cuts (2e2?)

• Lepton pT cuts
• pT1 gt thr1
• pT2 gt thr2
• pT3 gt thr3
• pT4 gt thr4
• mm- and ee- invariant mass cuts
• mZ1 lt thr5
• mZ2 gt thr6
• Four lepton invariant mass cuts
• thr7 lt mH lt thr8

leptons sorted in decreasing order of pT
mZ1 max(mmm-,mee- ), mZ2 min(mmm-,mee- )
35
Optimization of Selection Cuts (2e2?)

• Kinematic cuts are optimized simultaneously
  together with the isolation SpT threshold.
• Cut optimization performed using MINUIT by
  maximizing significance, defined by the
  Log-Likelihood ratio
• Cuts optimized independently for each Higgs mass.
• To exclude effects of limited MC statistics For
  each cut obtained from the automatic
  optimization
• Plot ScL vs cut value with all other cuts fixed.
• Assign final cut value by inspection, such that
  ScL is as close as possible to the maximum whilst
  retaining smooth variation of cut value as a
  function of mH.

where

36
Optimised Kinematic Cuts (2e2?)
37
?.BR.? After Each Cut
2e2?
x-axis categories Preselection, L1, HLT, 4
leptons, Vertex, Isolation, Lepton pT, Z
mass, Higgs mass
38
4 Lepton Invariant Mass Before/After Offline
Selection
mH130 GeV
mH200 GeV
Before offline selection
2e2?
After offline selection
39
Final Selected Events per fb-1 and NS/NB
2e2?
mH (GeV) 120 140 160 180 200 250 300 400 500
N signal for 10fb-1 1.9 11.7 7.8 8.7 36.4 29.1 19.4 18.0 9.6
N background for 10fb-1 1.5 2.0 2.0 4.0 16.2 13.6 4.1 3.7 2.6
40
Summary of Offline Selection for 4e Channel

• Longitudinal impact parameter significance for
  all electrons lt 13
• Transverse impact parameter significance of
  reconstructed Z() bosons
• lt 30 for highest mee-
• lt 15 for lowest mee-
• Isolation, required separately for each electron,
  cone size ?Rlt0.2
• Tracker isolation (?pTtracks)/pTe lt 0.1
• Hadronic isolation (?ETHCAL)/pTe lt 0.2
• Electron quality requirements
• Further cuts on track-cluster matching, cluster
  shape, HCAL/ECAL
• Kinematic cuts on lepton pT, mZ1, mZ2, m4e

41
Summary of Offline Selection for 4? Channel

• Find that only the following cuts are critical
• Isolation Tracker and calorimeter threshold
  applied to the least isolated muon
• Single pT threshold for each mass applied to all
  but the lowest pT muon
• Lowest pT muon already required to have pTgt7(13)
  in the barrel (endcaps)
• Four lepton invariant mass cuts
• Additional cuts (impact parameter, mm- inv.
  mass) do not significantly improve results.
• Cut optimization procedure similar to 2e2?
  analysis, but uses a minimization program named
  GARCON recently developed by the H?ZZ?4? group.

Calorimeter Isolation for least isolated muon
42
Evaluation of the ?Z()/?)(Z()/g) Background

• Systematic error on the no. on background events
  in the signal region enters into the significance
  calculation.
• Direct simulation of ?Z()/?)(Z()/g)?4l
  subject to the following uncertainties
• Theoretical uncertainties
• PDFs and QCD scale variations
• NLO and NNLO production cross-section
  uncertainties
• Relies entirely on existing SM constraints and
  theoretical knowledge
• Experimental uncertainties
• LHC luminosity
• MC modeling of detector response, material budget
  etc
• Energy scales (ECAL calibration) and resolution
• electron and muon reconstruction and kinematic
  selection efficiencies
• Electron and muon islolation efficiencies
• Such uncertainties are difficult to evaluate from
  first principles.
• More robust approach is to evaluate the size of
  the background directly using the LHC data.

43
Evaluation of the ?Z()/?)(Z()/g) Background
from Data

• 2 Approaches
• 1) Use single Z boson production
• Single Z bosons will be produced with a high rate
  at the LHC.
• Production cross-section will rapidly be measured
  directly to a high precision
• Can use ratio of production cross-sections for
  ?Z()/?)(Z()/g) and single Z production to
  evaluate the ?Z()/?)(Z()/g) background.
• Cancellation of luminosity uncertainties.
• Reduction of PDF and QCD scale uncertainties for
  low mH.
• Partial cancellation of experimental
  uncertainties.
• 2) Direct measurement through counting the number
  of events in the sidebands (i.e. excluding the
  signal peak) of the 4-lepton invariant mass
  distribution
• Full cancellation of all uncertainties except PDF
  and QCD scale uncertainties (not fully cancelled
  because may affect the shape of the m(4l)
  distribution).
• Disadvantage Limited by statistics of the
  background rate in the sidebands.
• Approach 2 is used here as the most robust
  solution.

44
Evaluation of ?Z()/?)(Z()/g) Background from
Sidebands
?L 9.2 fb-1
?L 5.8 fb-1
2e2?
2e2?

• Points represent a simulation of LHC data for the
  relevant integrated luminosities
• Total no. of events generated randomly from a
  Poisson distribution with mean total expected
  events from all processes (signal and
  background).
• For each event, 4 lepton invariant mass generated
  randomly according to the histogram formed from
  the sum of the MC distributions for signal and
  background.

45
Background Systematic Errors
Statistical error on background measurement from
data
Theoretical uncertainty on the ratio a
High statistical error at high mH due to low
statistics in sidebands due to hard lepton pT
cuts and large signal width.
2e2?
46
Background Systematic Errors Theory

• Systematic uncertainty from PDFs and QCD scale
  estimated using the MCFM event generator.
• 20 eigenvectors of the CTEQ6M PDFs varied by ?1?.
• QCD normalization and factorization scales varied
  independently up and down by factor 2 from
  nominal values ?R ?F 2mZ.

47
Significance Calculation

• Counting experiment significance, ScP
• Defined as no. of sigmas of a Gaussian
  distribution equivalent to Poisson probability of
  observing equal to or greater than NObs events,
  given ?B expected events
• An extended form of the ScP estimator is used
  which takes into account the systematic
  uncertainty on ?B.

48
Significance for 2e2? Channel
mH (GeV) 120 140 160 180 200 250 300 400 500
N signal at ?L for 5s 28.0 10.7 13.4 19.6 21.2 21.7 13.1 14.6 17.8
N back at ?L for 5s 21.4 1.8 3.5 9.1 9.4 10.1 2.8 3.0 5.3
49
Combined Significance for 30 fb-1
where
Without systematic Uncertainties
50
Combined Significance for 30 fb-1
Systematic uncertainties included
51
Higgs Mass Measurement from Gaussian Fit
mH140 GeV
mH200 GeV
mH500 GeV
Shown as fraction of true mass

• Statistical error on measurement of mH

• Measured Higgs mass from Gaussian fit for high
  statistics

2e2?
2e2?
52
Higgs Width Measurement from Gaussian Fit
2e2?
Shown as fraction of true width

• Direct measurement of width possible with
  ?statlt30 for mH?200 GeV

53
Higgs Cross Section Measurement Uncertainty
2e2?
Shown as fraction of expected no. of signal
events
54
Summary

• Standard Model Higgs boson with mass in range
  130mH500 GeV observable in the channel
  H?ZZ()?4l with gt 5s significance with 10fb-1 of
  integrated luminosity, excluding a 15 GeV gap
  close to mH170 GeV (40fb-1).
• If mass lies in the range 190mH400 GeV, 5s
  significance can be attained with 4fb-1.
• Size of ZZ/g background determined from data in
  sidebands with systematic uncertainty included in
  ScP significance calculation.

55
Aknowledgements

• H?ZZ?4e
• S. Baffioni, C. Charlot, F. Ferri, R. Salerno, Y.
  Sirois (LLR, France)
• N. Godinovic, I. Puljak (Split, Croatia)
• P. Meridiani (Rome and INFN, Italy)
• H?ZZ?4?
• S. Abdullin (FNAL)
• D. Acosta, P. Bartalini, R. Cavanaugh, A.
  Drozdetskiy, A. Korytov,
  G. Mitselmakher, Y. Pakhotin, B. Scurlock
  (Florida)
• A. Sherstnev (Cambridge)
• H?ZZ?2e2?
• D. Futyan, D.Fortin (UC Riverside)
• D. Giordano (Bari and INFN, Italy)

56
Backup Slides
57
Electron Experimental Systematic Uncertainties

• Material budget Change in amount of material
  traversed by electron before reaching the ECAL
  affects
• electron identification and selection
  efficiencies
• energy scale and resolution
• Material budget can be measured using single
  electron events, using the observed fraction of
  energy lost through bremsstrahlung, since energy
  radiated is proportional to material thickness
  traversed
• where pin and pout are the measured momenta at
  the innermost and outermost point on the GSF
  electron track.

where
58
Electron Experimental Systematic Uncertainties

• With electron statistics from single Z production
  corresponding to 10fb-1, can measure tracker
  material thickness to a precision better than 2

• 2 uncertainty shown to have almost no effect on
  electron reconstruction efficiency

59
Electron Experimental Systematic Uncertainties

• Electron reconstruction efficiency and energy
  scale can be controlled using tagged electrons
  from Z?ee events
• Select Z?ee events for which at least one leg is
  a golden electron (no bremsstrahlung), plus
  kinematic constraint on Z boson mass for second
  leg.
• Use second leg to estimate uncertainties on
  reconstruction efficiencies and on the energy
  scale.
• Systematic uncertainty on electron reconstruction
  efficiency and energy scale taken to be lt1

60
Muon Experimental Systematic Uncertainties

• Measure muon reconstruction efficiency from data
  to better than 1 precision
• Use sample of muon HLT triggers with pTgt19 GeV.
• Count no. of Z?2? events in the resonance of the
  inv. mass distributions built from
• HLT muon reconstructed muons
• HLT muon all tracks
• Ratio gives the efficiency.
• Can measure to better than 1.
• Efficiency of isolation cut measured by
  evaluating energy flow in isolation cones around
  random directions in Z?2? events.
• Can measure to better than 2.
• Uncertainty on pT resolution and pT scale
  evaluated using resonance peaks from Z?2? and
  J/??2? events to high precision.

61
Monte Carlo Event Generation Details

• In all samples, Z and W bosons forced to decay to
  e,m,t, with t forced to decay to e,??
• No forcing of b decays in Zbb and tt background
  events.
• In ?Z()/?)(Z()/g) and ?Z()/?)bb
  backgrounds, require m?Z()/?) gt 5 GeV
• Non-perturbative PDFs in the proton taken from
  CTEQ6 distributions
• Global QCD analysis combining all existing
  relevant deep inelastic and jet cross-section
  measurement results.
• QED final state radiation (internal
  bremsstrahlung) simulated by interfacing the
  event generators with dedicated software package
  PHOTOS.

62
True Significance of Local Event Excess

• Search for new phenomena in a wide range of
  parameter space - in this case narrow resonance
  in very broad range of invariant masses
• Problem of overestimating significance of a
  local discovery
• Need to reduce the significance according to the
  number of chances of getting it

63
Reconstructed Invariant Masses of ee- and mm-
mH 130 GeV
Electron pair
Muon pair
64
CP Nature of Higgs Shape of MZ Distribution

• Shape of MZ distribution depends on CP nature of
  Higgs
• Compare theoretical MZ distributions with result
  of convolution of reconstructed MZ distribution
  with efficiency of selection for MZ.
• Similar approach possible using cosq distribution
  of the angle between the planes containing the
  lepton pairs - important for MHgt2MZ

65
Choice of Trigger

• Natural choice from physics viewpoint is to
  trigger on Z
• (i) Take OR of 2e and 2m triggers
• Another possible choice is
• (ii) Take OR of 1e, 2e, 1m and 2m
• Consider fraction of events passing (ii) which
  fail (i). i.e. pass single triggers only
• For background, corresponds to close to half of
  events ? using 1e2e1m2m trigger results in
  almost twice as much background as 2e2m trigger
• For signal, using 1e2e1m2m rather than
  2e2m increases final no. of events after all
  offline cuts by lt1 for mHgt160 and lt5 for
  mHlt160.
• But this gain is offset by the fact that the no.
  of ZZ background events after offline cuts
  increases by a similar fraction.

Before offline cuts
After offline cuts
Conclusion use OR of double electron and double
muon triggers
66
Recovery of QCD Internal Bremsstrahlung

• At least one IB photon present in 40-45 of
  H?ZZ()?2e2? events
• At least one IB photon with pTgt5 GeV present for
  10-30 of events (increasing with mH)
• 2/3 emitted by electrons, 1/3 by muons
• Distinguish from other photons from the
  underlying
• event using tendency to be collinear with parent
  lepton.

• If gt1 reconstructed photons found within cone of
  size ?Rlt0.3 around any of the 4 reconstructed
  leptons, photon with smallest ?R is considered as
  an IB photon
• 4-momentum added to Z boson invariant mass prior
  to Z mass window cuts.