HLT etau selection

1
HLT etau selection
Signal, Susy Higgs gg?bbH with H???-?e-?e???jet
mA200 GeV/c2 tan?20
10000 signal events generated

• ?taujetlt2.4
• PTel gt14 GeV/c
• PT ?jetgt30 GeV/c

pythia preselection
2
qcd dijet sample statistics and preselection

• Egamma sample passing pythia preselection (CMS
  IN-2001/023)
• Total number of events after pythia preselection
  670 k events.
• Applied L1 e or etau preselection with loose cuts
• Low lumi (2x1033 cm-2s-1) High lumi (1034
  cm-2s-1)
• Et single electron gt 20 GeV Et single electron gt
  25 GeV
• Et (e,tau) gt (14 GeV,35 GeV) Et (e,tau) gt (14
  GeV,55 GeV)
• L1 preselected around 7 of events (depending on
  luminosity
• and pt bin)
• Remark Due to the egamma preselection, the rate
  calculation can
• only be performed if the trigger includes an
  electron trigger.
• for example single tau rate cannot be calculated
  without bias (see
• cms in-2001/023)

3
Purity at L1

• We need to guarantee a high L1 purity since the
  HLT trigger
• works by validating the L1 candidates.
• At high luminosity with (Etegt20,Et?gt62), the
  probablities (p)
• that an event that passes L2 will pass L25 tau or
  electron validation
• are
• p(real e passes L25 tau pixel L2 etau is
  true ) 0.63
• p(real taujet passes L25 ele pixel L2 etau
  is true) 0.18

We define the L1 tau as the most energetic tau
that is not colinear with the L1 electron
purity fraction of events where the real
electron and taujet are in the region that is
being triggered.
4
Choosing a working point _at_L1 L1034 cm-2s-1

• For a constant rate curve, decreasing the
  electron threshold cause an increase in the
  trigger efficiency. This is due to the fact that
  the electron Pt spectrum is sharper than the
  taujet Pt spectrum.
• From this plot take a point with 1kHz
  additionnal rate and see what happens after L2
  and L25 (applied on the electron and taujet)

5
Selection procedure L1034 cm-2s-1

• Treat electron stream
• and tau stream independenly
• Parameters
• L1SeleSingleIso1
• EPHitMatchePhiMin1 -0.025
• EPHitMatchePhiMax1 0.015
• EPHitMatchpPhiMin1 -0.015
• EPHitMatchpPhiMax1 0.025
• PixelTauMatchCone0.1
• PixelTauIsolCone0.40
• PixelTauSignalCone0.07
• PixelTauPtIsol1.0
• PixelTauPtMatch3.0

6
Rate and efficiency after L1,L2,L25 (L1034
cm-2s-1)
etau additionnal rate
single e rate
etau additionnal efficiency
single e efficiency
7
Spectrum recovery with etau trigger _at_ L25
Events triggered by the etau and not by the
single electron
Events triggered by the e
At high luminosity the etau trigger takes about
8 of the total e or etau trigger efficiency.
This is mutch lower at low luminosity since the
single electron trigger is already low (23 GeV)
8
Events repartition after L25 etau trigger

• Take events that passes the L2 etau trigger, see
  what is the relative efficiency of L25 as a
  function of the type of events
• class 0 events where the e and taujet are
  correctly assigned
• class 1 events where e is correctly assigned
  but not taujet
• class 2 events where taujet is correctly
  assigned but not the e
• class 3 events with e and tau not will assigned
  and not permutated
• class 4 events with e and taujet permutated

9
L3 strategy for etau (high lumi)
Preliminary

• Apply E/p cut on the electron candidate for the
  single electron and the electron in etau
• E/p lt 1.5 in barrel
• E/p lt 2 in endcap

Total L3 (e or etau efficiency) 35 etau
additionnal efficiency 3

• Need to evaluate background rate with
• Z???-
• tt-gtWWbb
• Wjet
• qcd dijet problem for the etau statistics (3
  events left for etau at low lumi).

10
Logical interface for combined trigger studies.

• Developping a code for combined trigger studies
  is not straight since there is no generic
  interface from which each trigger implementation
  derives.
• Playing with different logical trigger
  combination always need a rewriting of the code
  that implements the logic. (long cycle of
  rewriting, compiling, running). Need a way to
  specify the selection logic dynamically
• There is actually no ORCA code that gives you an
  bit pattern for the HLT trigger response. It
  Would be nice to have the whole HLT selection
  logic coded as it is discribed in daq/tdr

We have developed an interface into which part of
the HLT selection logic has been implemented and
can dynamically be changed via an xml file See
talk of Christophe.