Jets+met Triggers
SM Higgs boson search in the HZbb final state
Arnaud Duperrin (CPPM Marseille)
on behalf of D0 and CDF
TRIGGERS
1. Jets+MET Signals
2. Trigger Systems
3. Design
4. Historic
5. Data Trigger Efficiency
6. Performances
1. Data Sample
2. SM Backgrounds
3. The Multijet Background
4. Selection/Systematics
5. Results/Improvements
6. CDF
ANALYSIS
Alexandre Zabi (Oct. 04)
Bertrand Martin (Sept. 08)Florent Lacroix (Dec. 08)
Thomas Millet (May 07)
Christophe Ochando (Sept. 08)
Samuel Calvet (Sept. 07)
D0 France PhDs (on these topics)Fabrice Tissandier (Oct.
07)
2
1) Which Jets+met signals do we want to TRIGGER on?
gq ~,~
It is challenging:
3
2) Trigger System at D0 (online)
Full reco (offline)
4
Run IIb:
31032 cm-2 s-1
Why an upgrade of the trigger system in 2006 ?
data (min bias trigger) @601030 cm-2 s-1
Run IIa:
0.51032 cm-2 s-1
@2401030 cm-2 s-1
• new hardware (faster)
• new tools (ex MET)
• new design (ex: Oring)
5
3) Jet+MET trigger design: an example at L3
Signal ZH (mH=115 GeV)
25 GeV
(arbitrary normalization)
Result of the L3 design of Run IIb jets+met triggers for Higgs search cut rate to tape by ~50% while keeping trigger efficiencies constant (~85%)
Higgs and NP jets+met triggers are kept unprescaled up to highest luminosity
6
4) Jet+MET trigger: design historic
Fev. 2003 July 2003
• L1: CJT(3,5) : 3 TT with ET>5 GeV
• L2: MHT>20 GeV
• L3: at least 1 one jet, MHT>30 GeV, HT>50 GeV
June 2004v11 v12 v13
improved the triggers as function of the instantaneous luminosity increases
Fev.
2003
July.
2003June.
2004
MHT30
6% 23% of data) (RunIIa
3.01
fbdtL
FH
EM
CC
ECTower (TT)
Trigger Tower
2.02.0
(jets)pMHT
(jets)pHT
T
T
7
design historic
June 2004 July 2005v13 v14
June
2004
• JT1_ACO_MHT_HT
• JT2_MHT25_HT
38% 33%
(RunIIa) 11
fbdtL
July
2005
June
2006
Run IIb: Oring of several complex triggers very different from the first “MHT30” trigger
(and I am skipping a lot of the technical difficulties which went into these designs…)
• L1: MET>24 GeV and Jet Pt1>20 GeV and Jet Pt2>8 GeV and “no back-to-back jets” (noBB)
• L2: Pt1>20 GeV, MHT>20 GeV, HT>35 GeV, noBB
• L3: 2 Jets Pt>9 GeV, MHT>25 GeV, no BB (170o), (Jet1,MHT)>25o, MET>25 GeV
June 2006v15
Run IIb
• monjet+met
• dijet+met
• multijet+met
(RunIIb) 41
fbdtL
CC
L1JET CSWJT(1,8,3.2)
(June 2006)
8
TT calibration
Results: shown for L1 Jets and L1 MET
jetsTPL 1
TEL 1jets
TPL 1
DATA
MC
before after after
data/MC comparison for L1 objects entering in the
HZ triggers looks good after calibration (work in
progress)
5) Trigger Efficiencies
GEANT program does not simulate the D0 calorimeter response correctly
need to calibrate the response of the simulated trigger system with the data
Jet 1
Jet 2
QCD data/MC: 2 jets back-to back
Offline is ~OK
jetsTPo ffline
• TT calibration = bring the precision readout + shifting + smearing of TT energy to match data/MC
Two approaches:
1) Calibrate the online trigger simulator (called d0trigsim):
• get the jet+met trigger response
• takes the complex correlations between the objects (jets and MET)
• allows to study the systematics
9
Second approach: derive a standalone parametrization to “emulate” the jet+MET Higgs trigger response by calibrating objects directly and study possible remaining correlations (current choice for the analysis shown later)
6) Trigger Performances
+jets+MET triggers “emulation”
MHT
(both are data)
TE
HZ signal MC:
HZ Trigger efficiency: (for loose offline cuts)
•L1: 88%
• L1+L2+L3: 84%
• with un-calibrated d0trigsim: 91%
(+ complex Oring taken into account…)
Z+- +jets and W() + jets data: • equivalent to jets+met data from the calorimeter point of view• well understood signal and easy to collect (isolated muon trigger)
triggers
+jets+MET triggers
(both are data)
Example on how to parametrize:
term CSWJT(1,30,3.2) :
“at least one L1 jet with ET>30 GeV & ||<3.2”
mod TPoffline
Term efficiency
10
SM Higgs boson search in the HZbb final state
BR(Zl+l-) 3%
BR(Z)20% (3 neutrinos flavors)
ZHbb
WHlbb
ZHl+l-bb
with WH, ZH is the most sensitive channel at low mass
same final state than many NP particles (ex. sbottom, stop, LQ3)
(this search is also sensitive to W(l)H signal events when the lepton is not reconstructed represents 40% of the signal sample)
11
1) Data Sample 11.2
fbdtL
90% 81%2) SM Backgrounds
Z
b
b
q
q
W
b
b
Irreducible: Z()+jets
(800 pb)
reducible: W(l)+jets
(4500 pb)
(see Jean-Francois’s presentation on SM
backgrounds + heavy flavors “scale factors”)
12
3) The Multijet Background
Of the order of the milibarn (to be compared to signal cross section ~ 0.015 pb mH=115 GeV)
• jets:
(GeV) ET
Jets energy fluctuate MET
…but difficult to simulate (from theory and instrumental point of view) has to be evaluated with data (next slide)
multijet background contribution
jet 2
jet 1
MET
min(Jet,MET)
Selection:
• 2 or 3 jets Pt>20 GeV
• MET> 50 GeV
• min(Jet,MET), Aco veto…
13
QCD sample(>/2)
Signal sample(</2)
MET
SIGNAL QCD
M_trkPt
Jet 1
Jet 2MET
M_trkPt
(TrkPt, MET) is used to split the data in two samples: « QCD-like » and « signal-like »
R(jet 1, jet 2)
QCD
Z+jets
W+jets
in the “signal sample” at preselection level, the SM+QCD contributions (QCD obtained from data) shows a good agreement between data/MCSignal
x 500
14
4) SelectionBefor
eAfter
Dijet invariant mass
Signal (x10)
W+jetsW+saveurs lourdesZ+jetsZ+saveurs lourdesTopQCD
Signal (x500)
Neural Network b-tagging:
24 variables used:• dijet invariant mass(which is the most discriminant),• jets pT & ,• R(jet 1, jet 2), (jet 1, jet 2), etc…
Boosted decision tree (DT):
DT output
15
5) Systematics
• trigger efficiencies: 5.5%
• cross section: 6-16% (SM backgrounds), 6% (signal)
• HF fraction: 50%
• b-tagging efficiencies: 6%
• Luminosity: 6.1%
)bbVH)xBR(Hpσ(p
C.L. 95% @ excludedsection CrossRatio
(predicted by the “SM” Higgs)
At mH=115 GeV, Ratio=7.5 observed
(8.4 expected)
6) Results
most sensitive result for a low mass Higgs at D0
7) Improvements foreseen:
• lower the MET cut down to 40 GeV 15% more signal (including trigger efficiencies) work on trigger and QCD modeling
• combine with “single b-tagging” and separate 2 & 3 jets bins
• add an isolated track veto analysis
• jet resolutions improvements
• more luminosity!
16
8) CDF similar results (split by b-tagging categories + uses a NN to select the signal)
• Single Tagged category adds ~10% to sensitivity.
•Accept three jet events, where the 3rd jet is either a jet radiated off from a quark or a charged lepton => Adds sensitivity to WH->taunubb channel (hadronic jets = 30% of selected signal events)
•Multijet background shape and normalization are estimated from data => Multijet normaliztion uncertainty reduced to <20%.
•Jet energies are corrected using tracking information => Improves Dijet Mass resolution.
•Neural Network =>Improvement in signal acceptance with respect to cut-based selection
At mH=115 GeV, Ratio=7.9 observed
(6.3 expected)
CDF improvements in sensitivity of VH->MET+bb analysis in the course
of 2007-2008
win08
win08
17
Conclusion HZbb+ set one of the most stringent
limits on Higgs boson production cross-section among various
Tevatron searches many improvements still to
come combination with other channels (see Gregorio’s
presentation) Jets + MET triggers are challenging but provide access to very important search channels (not only Higgs but also to SUSY)
Undoubtly a very useful experience acquired at Tevatron in a challenging but important area which can be expanded at LHC experiments
with Fermilab, Manchester, Imperial College, among others