Recent Results in Susy Higgs Searches at DØ Jonathan Hays On behalf of the DØ Collaboration Fermilab Joint Experimental- Theoretical Seminar Friday, 12 th November 2010
Feb 24, 2016
Recent Results in Susy Higgs
Searches at DØJonathan Hays
On behalf of the DØ Collaboration
Fermilab Joint Experimental-Theoretical Seminar
Friday, 12th November 2010
Wine and Cheese Seminar 2
Outline
Introduction
Searches for Higgs + b-jets
tau final states (b)
b-jet final states (bbb)
Conclusions and Outlook
Wine and Cheese Seminar 3
Tevatron
5.2 fb-1
4.3 fb-1
Tevatron and the detectors continue to perform very well
~16 fb-1 expected by Oct 2014 Over 9.6 fb-1 delivered
Wine and Cheese Seminar 4
D-Zero
Wine and Cheese Seminar 5
Standard Model
Highly successful theory but:
No dark matter candidateNo gravityHierarchy and naturalness
problemsNo unification
Wine and Cheese Seminar 6
Supersymmetry
Solves naturalness problemLSP = dark matter candidate ?SupergravityGUT unification possible
hep-ph/9709356
Introduce new space-time symmetry between fermions and bosons
Wine and Cheese Seminar 7
MSSM Higgs Sector2 Higgs doublets
5 physical scalars: 3 neutral: h, H, A
2 charged: H±
A
hH
Totaltan()=30
MSSM
A. Djouadi, hep-ph:0810-2439
tree level two parameters: mA and tanβ
σMSSM ~ 2×Br×tan2β×σSM
Chance of discovery before SM sensitivity!
Radiative corrections large brings in dependence on other model parameters
Wine and Cheese Seminar 8
MSSM Higgs
→ b → 3b/b
bb→ 4b/bb
Enhancement to “down-type” fermions
BR(φ→bb) ~ 90%BR(φ→) ~ 10%
φ→clean signatures but low BR
bφ→breduced backgrounds added sensitivity at low mA
bφ→bbblarge backgroundhigh BR
φ = (h,H or A)
Wine and Cheese Seminar 9
MSSM Higgs
→ b → 3b/b
bb→ 4b/bb
Enhancement to “down-type” fermions
BR(φ→bb) ~ 90%BR(φ→) ~ 10%
φ→clean signatures but low BR
bφ→breduced backgrounds added sensitivity at low mA
bφ→bbblarge backgroundhigh BR
Wine and Cheese Seminar 10
Inclusive Searches
→
φ→
φ→
Wine and Cheese Seminar 11
Inclusive Searches
http://arxiv.org/abs/1003.3363v3
φ→
Tevatron combination
Wine and Cheese Seminar 12
Exclusive SearchesPublished results from DØ
bφ→bbφ→bbb
Phys. Rev. Lett. 104, 151801 (2010) Phys. Rev. Lett. 101, 221802 (2008)
Wine and Cheese Seminar 13
Search StrategyOptimise analysis based on expected limits with full systematics
In absence of significant discrepancy between data and background:Set limits in (almost) model independent waySet limits in benchmark SUSY scenarios
Combine results across channels for particular model choices
Wine and Cheese Seminar 14
Signal Modelling
Reweighted in pt and eta of spectator b-jet based on MCFM calculation
Important differences in kinematics when moving from LO to NLO
Use 5 flavour number scheme:
Generate gb→bh at LO in PYTHIAAcceptance cuts on the spectator b-jet
Wine and Cheese Seminar 15
Signal Modelling
Large enhancements to the couplings give large widths
Simulate widths using “narrow” samples and convoluting with Breit-Wigner
Radiative corrections have significant effect
Larger effect for bbb channelsLess significant for bττ
Wine and Cheese Seminar 16
b-jet identification
Several mature algorithms used:3 main categories:
- Soft-lepton tagging- Impact Parameter based- Secondary Vertex reconstruction
Wine and Cheese Seminar 17
b-jet identificationMeasure b & c efficiencies on b-jet enriched sample
Fake rate measured on multijet sample
Composition estimated from secondary vertex mass templates
MC and data differences
Data MC
b-tagged samples
Direct tagging Reweightwith TRFs
Tag rate functions (TRF) parameterise efficiencies and fake rates versus pt and eta
NIM A 620, 490 (2010)
Wine and Cheese Seminar 18
-lepton identification
TRK CALType 1
o
no TRK, but EM sub-cluster
TRK CAL
Type 2
³ 1 TRK
wide CAL cluster
Type 3
Hadronic decays categorised by decay mode
Leptonic decays – single isolated leptons
Neural network (NN) trained for each type to discriminate against jets
Efficiencies measured in clean Z sample
Wine and Cheese Seminar 19
Searches in tau final states4.3 fb-1 integrated luminosityCollected with single muon trigger
Dominant backgrounds:Z→ + jetstop pairsmulti-jet (QCD + W+jets)
Event selection: Single isolated muon Opposite sign had 1 loose b-tagged jet ( ε ~ 71%)
PreselectionNo b-tag
Complementary to φ→ and bφ→bbb
bφ→bτµτhad
Wine and Cheese Seminar 20
Searches in tau final statesTrain NNs to discriminate against top and multi-jet backgrounds
Final discriminant = geometric mean of 3 NN outputs
NN b-tagger suppresses Z+jets background
Combine all NNs into single discriminant
Wine and Cheese Seminar 21
bφ→bμhad limits
Tree level limit
Most stringent limit at low MA
4.3 fb-1 preliminary results
Limits set using “CLs” method
Wine and Cheese Seminar 22
Searches with b-quarks
5x more dataExtended mass range: 90-300 GeVLarger MC samples
New result with 5.2fb-1 data
Submitted to Phys. Lett. B arxiv.org:1011.1931
Expanded and improved treatment of systematics
- e.g. b-taggingRe-analyzed old 1fb-1 data set
Major improvements since previous 1fb-1 publication
Wine and Cheese Seminar 23
Searches with b-quarks
Very large multi-jet background
Challenging to model → data driven method
Multijet cross sections not well predicted → float normalisation
b → 3b/b
3 or 4 jets, 3 must be b-tagged
5.2fb-1 collected with jet triggers – making use of lifetime information
Kinematic likelihood (D) used to select best jet pairing, + cut to suppress background
Wine and Cheese Seminar 24
Background Modelling
MC correctionfactor
2 b-tagdata
3 b-tagbackground
Predict background shape from 2-tagged data with correction from MC
Add plot here...
2D correction: likelihood vs invariant mass
Wine and Cheese Seminar 25
Background Modelling:Sample composition
In 3-tag samplebbb ~ 47% bbj ~ 32%bbc ~ 17%ccj ~ 2%
Needed for MC correction factor
Estimated using MC fit to data over several b-tag operating points
Wine and Cheese Seminar 26
Background Modelling
Validate modelling in a signal poor region
“wrong” jet pair looks like background
Pick lowest likelihood pairing and select D < 0.12
Excellent agreement seen between model and data
Wine and Cheese Seminar 27
Kinematic Likelihood
Trained on jet-pairings
Two likelihoods: low mass MA < 140 GeV high mass MA ≥ 140 GeV
In each event select pairing with highest LH value
Cut on LH optimised considering expected limits with full systematics
LH > 0.65 for all mass points
Wine and Cheese Seminar 28
Kinematic Likelihood
Cut Cut
Projection of 2D distributions onto likelihood axis
Wine and Cheese Seminar 29
Mass distributionsDi-jet invariant mass distribution used as input for the limit setting
D > 0.65, background normalised to data
Wine and Cheese Seminar 30
Systematics
Background : normalisation included as nuisance parameter
Only consider variations in shape
Signal: dominated by b-tagging (15%-20%) and jet energy scale (2-14%) includes both rate and shape systematics
Wine and Cheese Seminar 31
Systematics: Fake-rate
An area of major improvement since 1fb-1 result
remeasured on hbb specific samples
Detailed approach to systematics
b-tagging SF SVT Template fit
Sample composition
Fake rate determinationFa
ke ra
teFa
ke ra
te
Wine and Cheese Seminar 32
Results
Wine and Cheese Seminar 33
Results
Small excess ~ 2.5σAfter trials factor ~ 2.0 σ
Wine and Cheese Seminar 34
SUSY Benchmark ScenariosFive additional parameters due to radiative correction
MSUSY (parameterizes squark, gaugino masses)Xt (related to the trilinear coupling At → stop mixing)M2 (gaugino mass term) (Higgs mass parameter)Mgluino (comes in via loops)
Two common benchmarksMax-mixing - Higgs boson mass
mh close to max possible value for a given tan
No-mixing - vanishing mixing in stop sector → small mass for h
Wine and Cheese Seminar 35
MSSM Scenario Limits
μ>0 suppressed production x BR – only set limits for μ<0
Wine and Cheese Seminar 36
OutlookStill large potential for improvements:
More data: 5 → 7+ fb-1
Improved b-tagging → 30% (bbb) yield
Improved analysis techniques e.g. Event based discriminants → 15-30% sensitivity
Wine and Cheese Seminar 37
Outlook: CombinationsCombine within channels – D0 + CDF – can be done in roughly model independent way
Combine across channels – generally requires picking a model
Aim for new D0 combination by Moriond with up to ~7fb-1
Preparations for Tevatron combination also underway
φ→
(φ→) + (bφ→b) + (bφ→bbb)
Wine and Cheese Seminar 38
Outlook: SM Contributions?
eg P. Draper et al. arXiv:0905.4721v2
Interpret SM limits within MSSM
Real potential to probe large region of MSSM Higgs parameter space
Wine and Cheese Seminar 39
ConclusionsInteresting time to be doing Higgs searches at the Tevatron!
Large data sets + sensitive analyses = discovery potential!
Wine and Cheese Seminar 40
Backup slides
Wine and Cheese Seminar 41
Mass distribution
Background normalised to data-signal
(S+B = D)
3-jet channels
Wine and Cheese Seminar 42
?
Wine and Cheese Seminar 43
Limit SettingUse modified frequentist
method “CLs”
Test statistic: negative poisson log likelihood ratio
Pseudo-experiments to extract likelihood distribution for B and S+B hypotheses
iiii
i
idi
bsbpd
ppLi
or
,!
)exp(
Systematics incorporated as Gaussian smearing in pseudo-experiments
Wine and Cheese Seminar 44
LLR Distributions
Background likeSignal like
CLb CLs+b