Impact Parameter Significance studies

Nick Edwards (University of Glasgow), Shih-Chieh Hsu (LBNL)

Introduction & Outline• Impact Parameter Significance definition:

• Useful in Electroweak analyses to reject heavy flavour backgrounds, as well as background from light quark backgrounds.

• Need to understand how well MC models the distribution observed in data, and derive smearing factors to reproduce data distribution in MC.

d0Sig = d0 / σ(do)

Data MC consistency

Central electron biased d0Sig wrt PVCentral muon biased d0Sig wrt PV

• Plots from T&P in Z mass peak 81 – 101 GeV with tight muons and Medium electrons with pT > 20 GeV, |η| < 2.5

• MC underestimates d0Sig – need to apply smearing• Strange spike in the electron distribution around 5 – this is also seen in the unbiased

distribution

Track Brem Refit• Shoulder of d0Sig distribution is due to Brem• Refitting track to account for energy lost to brem and recalculating d0Sig could solve this• Two approaches to Brem refitting:

• GSF (Gaussian Sum Filter) : The non linear effects of Bremsstrahlung are taken into account by modelling the energy loss at each material surface by sum of Gaussians which is then convolved with the original PDF describing the track parameters.

• DNA (Dynamic noise adjustment) At each material layer it performs a single parameter fit to determine if any energy was lost at that surface. If it detects that there was energy loss it will increase the the uncertainty of the track parameters (hence noise adjustment) by an amount corresponding to the measured energy loss.

Single lepton efficiencies: electrons• Select medium central

electrons following ZZ baseline selection (no isolation or d0). Apply d0 cuts and measure efficiency

• d0Sig has larger rejection efficiency for conversion electrons

• Brem refit d0Sig has better rejection efficiency for heavy flavour

• d0Sig and unbiased d0Sig lose more signal efficiency for the same background rejection compared to brem refit d0Sig

• Endcap rejection is larger for conversion compared to heavy flavour.• |DNA d0Sig| < 10 remains 99% signal efficiency with 15~20% rejection rate

• |DNA d0Sig| < 8 remains 98% signal efficiency with 20~30% rejection rate for different types of backgrounds source

Barrel Barrel

Barrel

Endcap

Shih-Chieh Hsu

IP Smearing

• Dataset: 2011 Egamma/Muon stream (B~E1)

• mc10 is normalized to 205pb-1• Select Zee/Zmumu within

81<mll<101 Region• Project d0Sig histogram• Extract the width of the

Gaussian core (+- 1σ)• Derive Smearing factor:• Smearing = sqrt(σdata

2 - σMC2)

Shih-Chieh Hsu

Data – MC smearing factors• d0Sig Biased Electron d0 error is underestimated such that width is larger than

for muon. However, the additional smearing factor 0.5σ is universal, with no obvious eta dependence

• d0Sig Unbiased: Combined muon unbiased d0Sig is from combined track instead of ID only track for unbiased d0Sig. For muons, the endcap smearing factor is smaller than barrel. Electron smear factor is not trivial

Eta Eta

Shih-Chieh Hsu

Data – MC smearing factors: Brem refit• Electron DNA refit d0Sig: (below) smearing

factor average is approximately 0.4. Flatter η distribution than for the un-refitted variables

• Electron GSF refit d0Sig: (right) MC has larger width than Data. Why does MC underestimate d0Err ?

Eta Eta

Shih-Chieh Hsu

Conclusions and Issues

• Clearly Impact parameter significance resolution worse in data than MC -> need smearing.

• Using brem track refitted parameters gives better background rejection / signal efficiency.

• Is it valid to use brem refitted track for d0 but not for other cuts (eta, isEM….)?

• Which is best refit algorithm to use?• Is it valid to apply smearing to d0 Significance rather

than to d0 ?

Backup

Loose Object Definition

Central Electrons• Author = 1 or 3• Use cluster energy, track direction

following egamma recommendation • Et > 15 GeV , |ηCluster|<2.47 • |Z0| < 10mm• Remove overlap with muons in cone

dR < 0.1• Overlapping electrons: remove the

one with lowest ET in cone dR < 0.1• Object quality (OQ & 1664 == 0)• Loose (at least 2 medium)

Forward Electrons• Author = 8• Et > 15 GeV 2.5 < |ηCluster| < 2.9• Overlap removal• Loose

Muons• STACO Tight.• MCP track Quality Cuts• Pt > 7 GeV• |η| < 2.5• |Z0| < 10mm

Forward Muons• STACO Tight• Pt > 15 GeV • 2.5 < |η|< 2.7

• https://twiki.cern.ch/twiki/bin/view/AtlasProtected/SMZZSummer2011

• Apply electron momentum smearing and muon momentum smearing and scaling to MC.

• <μ> Reweighting applied to MC using data weights for all 2011

• Data: 2011 B2 – E1 205pb-1• MC: mc10a, SMWZ d3pd

Loose Event Selection• Good Run List: data11_7TeV.periodAllYear_DetStatus-v13-pro08-

02_WZjets_allchannels.xml• Data: Remove events with Noise Burst and Data Integrity Error • Event Preselection:

– Require at least 4 tracks associated to primary vertex (to be discussed)– Trigger: 4mu: (EF_mu18_MG OR EF_mu18) . 4e: EF_e20_medium 2e2mu: either

trigger– Exactly four leptons as defined previously, >= 2 central leptons– Require at least one lepton Pt > 25 GeV

• ZZ formation:– Z01 : Choose closest same-flavour pair to the Z pole with at least one central

lepton, q0*q1 <= 0– Z34: Same-Flavour

• Z Mass Cuts: 66 GeV < MZ1 < 116 GeV, 20 GeV < MZ2 < 300 GeV• Z34 Opposite sign: q3*q4 <= 0

• Impact Parameter significance: d0PV / σ(d0

PV) < 10

DatasetsProcess dataset Xsec[pb] K-factor filterEffZZ -> 4l Pythiazz4l_3MultiLeptonFilterElecMu 0.07494 1.5 0.6235ZZ Herwig ZZ_Herwig 0.972 1.3 1

Top T1_McAtNlo_Jimmy 164.57 1 0.5562Z+X AlpgenJimmyZeeNp0_pt20 6.64E+02 1.26 1Z+X AlpgenJimmyZeeNp1_pt20 1.32E+02 1.26 1Z+X AlpgenJimmyZeeNp2_pt20 4.02E+01 1.26 1Z+X AlpgenJimmyZeeNp3_pt20 1.11E+01 1.26 1Z+X AlpgenJimmyZeeNp4_pt20 3.13E+00 1.26 1Z+X AlpgenJimmyZeeNp5_pt20 7.53E-01 1.26 1Z+X AlpgenJimmyZmumuNp0_pt20 6.64E+02 1.26 1Z+X AlpgenJimmyZmumuNp1_pt20 1.33E+02 1.26 1Z+X AlpgenJimmyZmumuNp2_pt20 4.04E+01 1.26 1Z+X AlpgenJimmyZmumuNp3_pt20 1.12E+01 1.26 1Z+X AlpgenJimmyZmumuNp4_pt20 2.90E+00 1.26 1Z+X AlpgenJimmyZmumuNp5_pt20 7.57E-01 1.26 1

WW McAtNlo_JIMMY_WpWm_enuenu 0.524 1 1WW McAtNlo_JIMMY_WpWm_enumunu 0.524 1 1WW McAtNlo_JIMMY_WpWm_enutaunu 0.524 1 1WW McAtNlo_JIMMY_WpWm_munuenu 0.524 1 1WW McAtNlo_JIMMY_WpWm_munumunu 0.524 1 1WW McAtNlo_JIMMY_WpWm_munutaunu 0.524 1 1WW McAtNlo_JIMMY_WpWm_taunuenu 0.524 1 1WW McAtNlo_JIMMY_WpWm_taunumunu 0.524 1 1WW McAtNlo_JIMMY_WpWm_taunutaunu 0.524 1 1

WZ McAtNlo_JIMMY_WpZ_lnull 0.15924 1 1WZ McAtNlo_JIMMY_WpZ_lnuqq 1.6889 1 1WZ McAtNlo_JIMMY_WpZ_lnutautau 0.07962 1 1WZ McAtNlo_JIMMY_WpZ_qqll 0.49836 1 1WZ McAtNlo_JIMMY_WpZ_taunutautau 0.03981 1 1