Susy Searches In Multilepton Final States

1
Susy Searches In Multilepton Final States

• Martin Niegel
• Karlsruhe Susy Group
• Karlsruhe Institute of Technology

2
Introduction outline
A

• Introduction
• Leptons in susy processes are mainly produced in
  ?20,?1 decays
• ?20,?1 are mainly produced in direct ?20?1
  pair production or in gluino, squark cascade
  decays
• decay topology is related to region in the
  m0-m1/2 space
• Outline
• Results of PTDR trileptons from ?20?1
  production at LM9
• Introduction of a multilepton susy search in
  inclusive mode

B
C
3
Trileptons from ?20?1 pair production

• Exclusive search for ?20?1? 3l no jets
• Tuned to LM9 (m0175,m1/21450,tanß50)
• m(?10)65GeV, m(?20,?1 )120GeV
• Backgrounds ZW,ZZ,WW,ttbar,Zjets,Wjets
• Soft leptons and small MET due to light ?10 and
  small mass difference ?20,?1 to ?10

D(m(?20,?10 ))
MET
invariant mass
pt of first lepton
4
Results for 30fb-1
Not a first day physics channel!

• trigger streams
• DiMuon stream (Ptgt 5 GeV) 75
• DiElectron stream (Ptgt17 GeV) 25

selection cuts NJets 0,Nlgt2,NOSSFgt0 Jets
itCone5, Jet_ptgt30GeV, etalt2.4 Muons
ptgt10GeV,etalt2.4 Electrons pt(1,2)gt17GeV,
ptgt10GeV,etalt2.4
Dilepton invariant mass
Discovery reach
5
Multilepton inclusive search
Inclusive any decay originating from SUSY
particles that gives leptons in the final
state

• Motivation
• larger cross section compared to exclusive
  trileptons (?20?1? 3l ) and can be use as first
  day physics analysis
• clean signature (Nleptons X) since
  uncertainties on leptons are smaller compared to
  JetMET
• (Maybe!) the selection can depend less on JetMET
  based cuts
• simultanious analysis of different Nleptons
  final states allows to constrain the model to a
  region in m0-m1/2 space
• becomes important in regions where JetEt MET
  is low( FocusPoint region)

Backgrounds ZJets,WJets, ttbar,WW,ZW,ZZ,...
Effective mass (SJetEtMET)
6
Multilepton signatures
Example of multilepton signatures from gluino
cascade decays
LM1
LM9
Cross section fb for multimuon final states on
generator level
7
Multimuon event properties Nmgt2 _at_LM9
CMSSW_1_6_12 CSA07DATASETS
Effective Mass
muon pt
effective mass
Muon_pt
DiMuon Minv
dimuon invariant mass
number of jets
NJets
Dimuon invariant mass GeV
8
Demo results _at_ LM9 1fb-1
NOT SOPHISTICATED!
NMuons 3
selection cuts NJets gt2,Meff gt400 Jets itCone5,
Jet1_ptgt100GeV, Jet2_ptgt60GeV, Jet3_pt3gt50
,Jet_etalt2.4 Muons TrackerMuon,TMLSTightID,
m(1,2,3)_ptgt10GeV,m_pt gt5GeV, etalt2.4
Dimuon invariant mass GeV
NMuons gt 4
NMuons 4
Dimuon invariant mass GeV
Dimuon invariant mass GeV
9
Variable selection
Challenge How to find the best discriminating
variables out of 1k different distributions ?
10
Variable selection
What are the properties of a perfect set of
variables ?
Rating example

• most discriminating power
• use rating algorithms
• least correlated
• use correlation Matrix
• least prone to uncertainties
• use smearing, toy experiments...
• least model dependent
• use FastSim scan in m0-m1/2

--------------------------------------- Rank
Variable  Separation --------------------------
-------------      1 var3     
3.834e02     2 var2       3.062e02
     3 var1       1.097e02    
4 var0       5.818e01 --------------------
-------------------
Correlations example
MVA (Multi Variate Analysis Toolkit) gives access
to a bunch of statistical algorithms, variable
rating and correlation matrix and was implemented
in /PhysicsTools by Christophe Saout (KIT)
Talk on MVA Tuesday 940h C.Saout
11
Conclusion

• 5s can be observed with 30fb-1 in Trileptons
  from direct Neutralino/Chargino pair production,
  but is not a first day physics channel
• Inclusive multilepton search benefits from larger
  cross section compared to exclusive search and
  has a cleaner signature (leptons) compared to Jet
  MET based analysis, since leptons are less
  prone to uncertainties
• Simultanious analysis of different multilepton
  final states at once can constrain the model to a
  region in m0-m1/2-space
• First analysis for inclusive multimuon search
  gives already 5s for 1fb-1 with not
  sophisticated selection
• Next steps use MVA toolkit to improve selection,
  take uncertainties into account and try to make
  selection more model independend