Performance of the PtmissPlusLeptonsFilter

1
Performance of the PtmissPlusLeptonsFilter

• Pauline Gagnon and Siva Subramania
• Indiana University

developed for the invisible Higgs analysis in HZ
channel with Z ? l l
2
Z0H0? ll- invisible initial analysis
strategy(Fortran-based ATLFAST,
ATL-PHYS-PUB-2005-011)

• Trigger 1 or 2 prompt leptons
• Preselection
• large missing pT
• 2 leptons of opposite sign and same flavour
• loose cut on Z mass
• Final selection
• Use 8 discriminative variables in a likelihood
  analysis to extract signal

3
Invisible Higgs decay in HZ channelall MC
samples generated with Pythia

• Signal
• H ? ?? and Z ? l l
• Large missing pT
• Two prompt leptons forming a Z
• Main backgrounds
• ZZ ? l l ??
• Z inclusive with Z ? l l and jets lost in the
  beam pipe
• tt ? b l ? b l ? with bb jets lost in the beam
  pipe
• W W ? l ? l ?
• W Z ? l l l ? with one lepton lost in the beam
  pipe
• ZZ ? l l ??

4
Most background were rejected at preselection
level in ATL-PHYS-PUB-2005-011
Create a filter to emulate preselection cuts
ptmiss cut, 2 leptons reconstructing a Z with
loose mass cut
5
Missing pT in reconstruction code

• Missing pT - ? (visible transverse energies)
• Isolated photons, electrons and muons
• Jets, b-jets and c-jets
• Clusters not accepted as jets
• Non-isolated muons not added to a jet
• Transverse energies in cells not used elsewhere

6
Missing pT at generator level

• Vectorial sum of transverse momenta of
• Invisible particles (stable Higgs, neutrinos)
• Lost particles those falling outside the
  detector acceptance region
• Calorimeter acceptance ? lt 5.0

Ptmiss calculated differently in filter and full
reconstruction chain So we are not comparing the
same object
7
Second handle in filter leptons

• Two leptons within the tracker acceptance region
• One or both leptons must pass the lepton trigger
  requirements
• Two leptons of same flavor but opposite sign
• The reconstructed mass must be close to mZ

8
Filter versus preselection cuts

• Filter cuts looser cuts
• 2 leptons with ?lt 2.7
• Trigger requirements
• - 1 e with pT gt 18 GeV or
• - 2 e with pT gt 23 GeV or
• - 1 µ with pT gt 8 GeV or
• - 2 µ with pT gt 13 GeV
• Missing pT gt 50 GeV
• mZ 55 GeV

• Preselection cuts
• 2 leptons with ?lt 2.5
• Trigger requirements
• - 1 e with pT gt 20 GeV or
• - 2 e with pT gt 25 GeV or
• - 1 µ with pT gt 10 GeV or
• - 2 µ with pT gt 15 GeV
• Missing pT gt 90 GeV
• mZ 50 GeV

9
Comparison with ATLFAST 10.5.0
Conclusion no significant event loss at the
filter level w.r.t. ATFAST with 10.5.0 or 11.0.41
(mZ cut)
10
Cross-check no statistical difference between
filtered and unfiltered events at AOD level
11
CPU time savings to simulate 100 fb-1 of data
From 8360 CPU years down to 20 CPU years
12
Cut flow on ATLFAST AOD after filter with mZ cut
jet and b-tag values not yet reliable in ATLFAST
13

• Reference plots done with ATLFAST AOD for
  unfiltered and filtered before selection
• for HZ events (plots are superimposed)

14
Reference plots after selection unfiltered and
filtered HZ events(stacked histograms)
pT lepton 2
pT lepton 1
Cos between leptons (3D)
Cos between leptons (2D)
15
Reference plots after selection unfiltered and
filtered HZ events(stacked histograms)
transverse mass
missing pT
Cos pTmiss-pT lepton 1
mZ
16
Filter performance with full AOD

• Tested with 1000 fully reconstructed events for
  HZ ???l l and for ZZ ? l l ??
• More checks in progress with ttbar samples
• All events preselected on fully reconstructed AOD
  events were retained by the filter

This filter should be used when producing signal
and background samples for HZ ??? l l analysis
17
Comparison of cut flow for fast and full
reconstruction
18
Conclusion

• This filter significantly reduces the amount of
  events to be reconstructed
• There is no significant loss of events for signal
  and main backgrounds w.r.t. events preselected
  with ATLFAST and full simulation AOD
• Filter has been tested with releases 10.5.0 and
  11.0.41 yielding consistent results
• This filter should be used for official signal
  and background production for invisible Higgs in
  HZ channel
• Filter is going to be in release 11.5.0