Background Studies

1
Background Studies
2
Background TPC

• Simulated 2000 bunch crossings (BXs) of beam
  background
• For TPC, conservatively take drift velocity to
  be 4 cm ms-1
• Therefore fill TPC with 150 BXs of background
  shifted in z
• First order attempt to merge unresolvable hits
• Superimpose on fully-hadronic top-pair events at
  500 GeV

150 BXs of pair background
3

• Large fraction of hits from low energy
  electrons/positrons
• from photon conversions
• Form tight helices, micro-curlers, along
  length of TPC
• Background concentrated on relatively few TPC
  readout pads
• Developed PatRec software to identify and remove
  micro-curlers

150 BXs of pair background
4

• Effective removal of large fraction of
  background hits

• By eye clear that this should be no problem
  for PatRec

5

• Superimpose 150 BXs TPC background on
• For 100 events, NO loss in track-finding
  efficiency observed
• Similar story for 3x nominal background
• Believe this is a clear demonstration of the
  robustness of a TPC
• operating in ILC beam conditions

6
Background VTX

• Background in VTX detector complicated by
  assumptions for
• Si pixel readout rate
• IF one assumes single BX tagging capability then
  background is
• not an issue
• For ILD studies conservatively assume 30 ms /
  125 ms integration
• times for VTX layers (0,1) and (2,3,4,5)
  respectively
• Therefore VTX integrates over 83/333 BXs
• Superimpose on fully-hadronic top-pair events at
  500 GeV

• Also consider finite cluster size of
• background hits (10 pixels)
• Significantly increases occupancy

7
Background VTX - fake tracks

• Combinatorics produce fake ghost tracks
• In addition to some real electron/positron
  background tracks
• Large combinatoric background challenges pattern
  recognition
• Reconfigured current algorithm (not ideal)
• From 83/333 BXs overlayed on
  
• reconstruct 34 ghost tracks/event (1/3
  are genuine)
• Rejected by requiring at least 1 SIT hit or gt10
  TPC associated hits

34/event
1/event
Left with 0.5 GeV per event (mixture of real
tracks/combinatorics)
8
Background VTX tracking efficiency

• Two effects potentially reduce tracking
  efficiency
• VTX pattern recognition
• Occupancy - assume physics hits next to
• background clusters
  lost
• Use parameterisation of occupancy/cluster size
  to kill physics hits
• superimpose 83/333 BXs VTX
  background
• apply SIT/TPC BX-tagging requirements

NOTE

• Care needed in interpreting efficiency results
• Will get different results depending on
  denominator
• e.g. if calculate efficiency for tracks
  with gt100 TPC hits,
• the efficiency will be 100
  

• Here we show results for
• all charged particles with pT gt1 GeV and
  NVTXNSIT gt 4
• as above, but for charged particles which reach
  the TPC
• (i.e. in MC leave at least 1 TPC hit)

9

• Background mainly affects reconstruction of low
  pT tracks
• pT gt 1 GeV efficiency reduced by 0.1
• For charged particles which reach TPC (i.e.
  dont decay/interact)
• pT gt 1 GeV efficiency 99.9 in presence of
  background

Nominal ILC background not a major problem for
ILD concept
10
Impact in a physics analysis

• Given limited time it has not been possible to
  superimpose full
• 83/333 BX in VTX, 150 BX in TPC and 1 BX in
  SIT on physics events

• CPU resources too large with current pattern
  recognition code
• TPC track finding shown not to be an issue
• Ghost tracks unlikely to be important for ZH ?
  mmX
• Possible loss of hits due to occupancy likely to
  be main effect
• could degrade momentum resolution fast to
  simulate

Barely noticeable
11
Background flavour tagging efficiency

• Simulated effect of VTX occupancy on flavour tag
• expected to be main contribution due to
  LCFIVertex track quality cuts

• Essentially same performance, not clear if
  anything other than
• statistical differences

12
Background summary

• Much progress since Tsukuba (thanks to IDAG
  impetus)
• Factorised problem into several parts
• TPC tracking appears to be very robust for ILC
  background
• Background in VTX does not cause significant
  problems
• 1 background track/event ltpTgt 0.5 GeV
• will improve with better pattern recognition
  software
• Efficiency losses are small even when accounting
  for finite
• cluster sizes
• Impact on physics (ZH) not fully studied but on
  basis of
• the studies above, no degradation expected
• Final comment
• Impact of background depends strongly on
  assumptions
• We have taken a conservative approach for VTX
• readout times (and TPC drift velocity)

13
ZH Recoil Mass
14
Higgs Recoil Mass

• New signal samples with fixed beamstrahlung
  spectrum

• Decided not to regenerate background samples
• Reweighted old samples using new/old
  distributions
• of E(e) and E(e-) of colliding particles
• Carefully checked tested using old ZH sample

• Select events using only information from
  di-lepton system

15
Also studied inclusion of Bremsstrahlung

• Fairly simple algorithm
• Does a reasonable job, but not optimised for
  recoil mass
• Significantly increases height of recoil mass
  peak
• Improves resolution above peak
• But, degrades resolution below peak (ECAL
  resolution worse than track)

16
Model independent results
s(mH) 32 MeV

• For Model Dependent analysis

s(mH) 27 MeV
Relation to detector performance

• This is a benchmark analysis for momentum
  resolution
• Width of mmX recoil mass peak
• 560 MeV for perfect resolution
• 650 MeV after reconstruction
• For ILD momentum resolution, luminosity
• spectrum still dominates !
• 560 MeV vs 330 MeV

ILD momentum resolution well matched to ZH
requirements
17
Physics Updates
18
Higgs Branching ratios

• Determine BR(H?bb), BR(H?cc), BR(H?gg) from
  Higgs-strahlung events
• Previously noted SiD had much small error than
  ILD in qqcc channel
• (for essentially same flavour tagging
  performance)
• Still not understood SiD/ILD working to
  understand discrepancy
• In addition, we have updated the nncc analysis
• Previously selected nncc and then used cross
  section/total cross section
• from M.I. recoil analysis to obtain BR
• This is incorrect as background from ZH? nnbb
  was taken to be SM value
• Now simultaneously extract all H?cc, H?bb, and
  H?other hadronic BRs
• In addition, smaller total cross section errors
  from recoil analysis

19
Physics Updates tau polarisation

• In principle, ILD well optimised for tau spin
  analysis
• 55 mm2 ECAL granularity for particle
  separation
• large BR2 for track/photon separation
• large R for photon/photon separation

• Original LoI analysis was rather basic
  incomplete
• Moved to SiD acceptance
• Moved to ANNs for decay mode ID
• Moved to optimal observables for Pt

Excellent performance
(not including a1)
Very large improvements c.f. LoI
20
And finally, Top at vs 500 GeV (Update)

• In addition to

• Have now added fully-hadronic vertex-charge based
  

(still relatively unsophisticated)
20 selection eff. 80 correct charge