Transcript Standard Model Higgs Searches at LHC

Standard Model Higgs
Searches at LHC
Suyong Choi
Korea U
SM HIGGS PRODUCTION AND
DECAY
SM Higgs Production Cross
Sections at 7 TeV
SM Higgs Production Cross
Sections at 14 TeV
Branching Fractions
SM Higgs 𝜎 × 𝐵𝑟
• Sensitivity depends
on
– 𝜎 × 𝐵𝑟
– Backgrounds
– Mass resolution
More info in https://twiki.cern.ch/twiki/bin/view/LHCPhysics/CrossSections
SM Higgs Search Channels
• 𝑀𝐻 < 130 𝐺𝑒𝑉
– 𝛾𝛾 – good mass resolution
– 𝑍𝑍 → 4ℓ - clean, good mass resolution
– 𝐻 → 𝑏𝑏, 𝜏 + 𝜏 − - not clean, worse mass resolution
• 130 < 𝑀𝐻 < 180
– 𝑊𝑊 – statistics
• 𝑀𝐻 > 180 𝐺𝑒𝑉
– 𝑍𝑍 → 2ℓ2𝜈 – statistics, clean
– 𝑍𝑍 → 4ℓ - clean, good mass resolution
• Overall, they are very complicated analyses
SM HIGGS SEARCHES AT CMS
AND ATLAS
CMS SM Higgs Channels
ATLAS SM Higgs Channels
𝐻 → 𝜏τ
VBF selection
Boosted selection
1 jet pT>150 GeV
𝐻 → 𝜏τ
𝐻 → 𝑏𝑏
• W/Z+H
𝐻 → 𝑏𝑏
𝑯 → 𝜸𝜸
𝐻 → 𝛾𝛾
• Event categories divided into
– 2 classes where the smallest 𝑅9 of two
photons is less or greater than 0.94
– 2 classes where the largest 𝜂 is in endcap
or barrel
– Total of 2x2=4 classes
Mass resolution for 𝑀𝐻 = 120 𝐺𝑒𝑉
𝐻 → 𝛾𝛾
Excess: (1.7 ± 0.8) × 𝜎𝑆𝑀
SM signal x5
Consistency
• P-value - Probability that background to produce
fluctuation as large as observed
2.3
@123.5 GeV
𝐻 → 𝛾𝛾 Upper Limit
Data disfavors Higgs in 127 – 131 GeV @ 95% CL
ATLAS 𝐻 → 𝛾𝛾Mass resolution
1.7 GeV
ATLAS 𝐻 → 𝛾𝛾
• 114 – 115, 135-136 GeV excluded @
95% CL
∗
𝑯 → 𝒁𝒁 → 𝟒ℓ
𝐻→
∗
𝑍𝑍
→ 4ℓ
• ZZ selection
– A second lepton pair:
𝑚𝑍2 > 12 𝐺𝑒𝑉
– 𝑚4ℓ > 100 𝐺𝑒𝑉 for 4e,
4
– Two sets of cuts for
low-mass and highmass Higgs
• Signal efficiencies
Channel
𝒎𝑯 = 𝟏𝟗𝟎 𝑮𝒆𝑽
𝒎𝑯 = 𝟒𝟎𝟎 𝑮𝒆𝑽
4e
49%
59%
2e2
61%
71%
4
78%
82%
𝐻→
∗
𝑍𝑍
→ 4ℓ
• Higgs mass resolutions
Channels
𝒎𝑯 = 𝟏𝟓𝟎 𝑮𝒆𝑽
𝒎𝑯 = 𝟏𝟗𝟎 𝑮𝒆𝑽
4e
2.7 GeV
3.5 GeV
2e2
2.1 GeV
2.8 GeV
4
1.6 GeV
2.5 GeV
𝐻→
∗
𝑍𝑍
→ 4ℓ
• Backgrounds
72 observed
67.1 ± 6.0 exepected
– Reducible - 𝑍𝑏𝑏, 𝑍𝑐 𝑐,
𝑍𝑗𝑗
– Irreducible - 𝑍𝑍
– All derived from data
Theory: 27.9 ± 1.9 𝑓𝑏
𝐻→
∗
𝑍𝑍
→ 4ℓ low mass region
• 13 events observed
• 9.5 ± 1.3 expected
Channels
Expected
Observed
4e
1.7
3
2e2
4.5
5
4
3.3
5
• No significant excess
𝐻→
∗
𝑍𝑍
→ 4ℓ
Limits from 𝐻 →
∗
𝑍𝑍
→ 4ℓ
340~465 GeV
134~158 GeV
180~305 GeV
expected exclusion: 130-160 GeV, 182-420 GeV
ATLAS 𝐻 → 𝑍𝑍 → 4ℓ
71 events observed
629 events expected
Below 180 GeV,
8 events observed
9.31.5 events expected
2e2μ events (m=123.6 GeV, m=124.3 GeV), one 4μ event (m=124.6 GeV)
ATLAS 𝐻 → 𝑍𝑍 → 4ℓ
ATLAS 𝐻 → 𝑍𝑍 → 4ℓ
135 – 156 GeV
excluded
181-234 GeV
excluded
255-415 GeV
excluded
∗
𝑯 → 𝑾𝑾 → 𝟐ℓ𝟐𝝂
Further selections
• mass-dependent selection
– 𝑝𝑇ℓ , 𝑚ℓℓ , Δ𝜙(ℓℓ), 𝑚 𝑇
Yields after signal selection
– Experimental uncertainties only
– Signal efficiency uncertainty ~ 20%
– Background uncertainty in signal region ~
15%
𝑯→
∗
𝑾𝑾
→ 𝟐ℓ𝟐𝝂 Limits
129-270 GeV Excluded @ 95%CL
127-270 GeV expected exclusion
ATLAS 𝐻 → 𝑊𝑊 → 2ℓ2𝜈
ATLAS 𝐻 → 𝑊𝑊 → 2ℓ2𝜈
ATLAS 𝐻 → 𝑊𝑊 → 2ℓ2𝜈
• 2.05 fb-1
110 events observed
9110 expected
If Higgs of certain mH existed
𝐻 → 𝑊𝑊 → 2ℓ2𝜈
• 145 – 206 GeV excluded @ 95% CL
– Excpected exclusion: 134 – 200
∗
𝑯 → 𝒁𝒁 → 𝟐ℓ𝟐𝝂
𝑯→
•
•
•
•
∗
𝒁𝒁
→ 𝟐ℓ𝟐𝝂
Dilepton trigger
Veto events with 3rd lepton
Cuts to reject Fake Missing ET
Final selection
– MET cut – mass dependent
– MT
Backgrounds
• MET modeling using
𝛾 + 𝑗𝑒𝑡𝑠 events
– reweighting according
to n-jets, boson pT
– Less reliance on MC
simulation
• Data driven methods
to estimate nonresonant backgrounds
– Top pair, single top, WW,
W+jets, 𝑍 → 𝜏𝜏
𝑯→
∗
𝒁𝒁
→ 𝟐ℓ𝟐𝝂
𝑯→
∗
𝒁𝒁
→ 𝟐ℓ𝟐𝝂 Limits
270-440 GeV excluded at 95% CL
CMS COMBINATION
Expected exclusion: 117 – 543 GeV
Global p-value 1.9 with LEE in 110~145 GeV
0.6 with LEE in 110~600 GeV
CMS Combined Higgs Exclusion
Limits
ATLAS COMBINATION
RESULTS
Consistency with Background only
hypothesis
• 3.6 excess
– 𝛾𝛾: 2.8
– ZZ*: 2.1
– WW*: 1.4
• With LEE
– 3.6→2.3
– 7% to observe
excess in 𝛾𝛾
– ~30% to observe
excess in ZZ
• SM expectation
is 2.4 for
126 GeV Higgs
1.9x10-4
Combined ATLAS SM Higgs
Exclusion Limits
95% exclusion limits:
112.7 - 115.5 GeV
131 – 237 GeV
251 – 453 GeV
Expected 95%CL
exclusion:
124.6 – 520 GeV
99% exclusion limits:
131 – 230 GeV
260 – 437 GeV
Summary and Outlook
• Atlas and CMS data narrowed the
allowed mass range for SM Higgs
– ATLAS : 115.5 – 131 GeV
– CMS : 114 – 127 GeV
• 20 fb-1 more data per experiment in
2012 allows 5 observation per
experiment at mH=125 GeV
BACKUP
Dataset
Lumi
Uncertainty
4.5%
Good data up to 4.7 fb-1 used in the updated analyses
Backgrounds
WW Selection event yields
𝐻 → 𝛾𝛾
• Background modeling
– MC simulation of background was not used
for background estimation, but in
agreement with data
– 30% non-prompt photons
– 5th order Bernstein polynomial fitted to the
100 < 𝑚𝛾𝛾 < 180 𝐺𝑒𝑉
• Maximize sensitivity
𝐻 → 𝛾𝛾
• Signal
– 110 < 𝑚𝛾𝛾 < 150 𝐺𝑒𝑉 in 5 GeV steps (9 mass
points)
– POWHEG NLO + PYTHIA
– Higgs 𝑝𝑇 reweighted to NNLL+NLO
• Using HqT program
• Fine corrections to photon energies
– Intercalibration
– Transparency corrections
– Improves resolutions by 10%
𝐻 → 𝛾𝛾
• Diphoton trigger
– Asymmetric ET thresholds
– complementary photon quality selections
– 100% trigger efficiency
• Photon energy corrected for conversions
upstream of Electromagnetic calorimeter
– Boosted decision tree regression trained on
MC samples
𝐻 → 𝛾𝛾
• Vertex location
– Mean number of pp interactions ~ 9.5
– RMS spread in beam direction ~ 6 cm
– 10mm accuracy in vertex location ensures that energy
resolution is not spoiled
• Identifying the correct vertex
– Kinematic properties of tracks emerging from the vertex
and their correlation with diphoton kinematics
• Sum of track 𝑝𝑇2 , momentum balance
– Converted tracks point to vertex
• 3% gain in efficiency
𝐻 → 𝛾𝛾
• Photon kinematic selection
– 𝑝𝑇1 >
𝑚𝛾𝛾
3
, 𝑝𝑇2 >
𝑚𝛾𝛾
4
– 𝜂𝛾 < 2.5, excl. barrel-endcap transition
• Backgrounds
– Irreducible 𝛾𝛾
– Fakes: 𝛾 + 𝑗𝑒𝑡, dijet
𝐻 → 𝛾𝛾
• Photon isolation
– Energies in Ecal and Hcal – affected by pile up
• Estimate effect of pileup in the event by average energy
density away from jets
– charged tracks around the photon candidate – fake
vertex allows non-isolated photon to appear
isolated
• Calculate track isolation w.r.t. vertex that maximizes it
• Photon quality
– H/E
– Transverse width of a photon shower
– Electron track veto (E/p)
𝐻 → 𝛾𝛾
• Dividing photon candidates
– Different S/B for photons of different criteria
– Barrel vs Endcap
• Barrel photon has less QCD background
– 𝑅9
• Energy in a 3x3 crystals around highest energy /
supercluster energy
• Photons with large 𝑅9 have less probability to
have converted
𝐻 → 𝛾𝛾
• Photon ID efficiencies
– Measured using 𝑍 → 𝑒 + 𝑒 − , excluding track
veto eff.
Systematic Uncertainties in 𝐻 → 𝛾𝛾
𝐻→
•
•
•
•
∗
𝑍𝑍
→ 4ℓ
3 channels – 4e, 4, 2e2
Covers 110 – 600 GeV
Used 4.7 fb-1
Triggers
– Dilepton triggers with asymmetric thresholds
of pT>8, 17 GeV
𝐻→
∗
𝑍𝑍
→ 4ℓ
• Offline
– Electrons pT>7 GeV, 𝜂𝑒 < 2.5, (90% for 𝑝𝑇𝑒 ≈
20 𝐺𝑒𝑉)
– Muons pT>5 GeV, 𝜂𝜇 < 2.4, 98% efficient
– Small impact parameter significance<4
– Z1: lepton pair with mass closest to mZ and
50 < 𝑚𝑍1 < 120 𝐺𝑒𝑉
𝑯→
∗
𝑾𝑾
→ 𝟐ℓ𝟐𝝂
• 2 leptons + MET
– ee, e, 
– 1 or 2 high pT leptons in the trigger
• 97~99% efficiency for signal of mH=160 GeV
– 0, 1, 2 jet categories considered
Offline Selection
• Offline
– Lepton pT 20 GeV, 10(15) GeV for e(ee,), Consistent with
coming from Vertex
– Jets 𝐸𝑇 > 30 𝐺𝑒𝑉, 𝜂 < 5
– Projected missing ET>20(40) e(ee,)
– Azimuthal opening angle dilepton-leading jet < 165 degrees
(ee,)
– Dilepton mass cut
• Remove low mass resonances, Z
– Reject events where jets tagged with soft leptons or large
impact parameter tracks
• Remove top events
– Reject events with 3rd isolated lepton
• Remove ZZ, WZ
– Identify converted photons to reject 𝑊𝛾
Background estimation
• Mostly data driven
– Apply antiselection, then extrapolate to
signal region
– W+jets, QCD multijets
– 𝑡𝑡, 𝑡𝑊
– 𝑊𝑊 – select events 𝑚ℓℓ > 100 𝐺𝑒𝑉
– Statistics of control sample limits
background estimate error
WW+0 jet baseline selection
WW+1-jet baseline selection
𝑯→
∗
𝑾𝑾
→ 𝟐ℓ𝟐𝝂
ee + 𝜇𝜇
0 jet
1 jet
𝑒𝜇