Efficiencies and fakes in muon trigger chains M

Efficiencies and fakes in muon trigger chains M. Bianco, E. Gorini, M. Primavera, A. Ventura INFN Lecce and University of Salento Muon Trigger Signature Group Meeting, February 3 rd, 2011

Outline Datasets, references, chains Efficiencies Fake with Tag&Probe, data vs MC probability: definitions Results on fake probabilities for HLT ◦ Comparison among various fake definitions ◦ Comparison with different offline reference ◦ Comparison on different streams Conclusions 03/02/2011 A. Ventura - Fake probability in muon trigger chains 2

Datasets, chains, references Runs from 166786 (H 1) to 167776 (I 2) Streams analyzed: ◦ physics_Muons ◦ physics_Jet. Tau. Etmiss Chains of interest ◦ mu 13_tight ◦ mu 13_MG_tight ◦ mu 20 ◦ mu 40_MSonly (both L 2 and EF, all unprescaled) Offline algorithms used ◦ Muid (SA and CB) ◦ Staco (SA and CB) 03/02/2011 as references: A. Ventura - Fake probability in muon trigger chains 3

Efficiency definition Primary vertex with ≥ 3 tracks, |z|<150 mm Muid. CB offline reference Reference offline muons with requirements https: //twiki. cern. ch/twiki/bin/view/Atlas. Protected/MCPAnalysis. Guidelines. Rel 15 Tag ◦ ◦ ◦ & Probe selection: 2 opposite sign reference muons, reciprocal DR>0. 5 |m(m 1 m 2)–m. Z|<15 Ge. V Tag if p. T>10 Ge. V and an EF_mu 1 X object within DR<0. 15 Probe muon provides efficiency if it matches a trigger object within DR<0. 5 03/02/2011 A. Ventura - Fake probability in muon trigger chains 4

L 2_mu 13_tight : Data vs MC Tag & Probe efficiency on data (Muons stream) and on MC (mc 10_7 Te. V. 106047. Pythia. Zmumu_no_filter. merge) Barrel Endcaps MC/Data = 1. 017 ± 0. 012 03/02/2011 MC/Data = 0. 992 ± 0. 005 A. Ventura - Fake probability in muon trigger chains 5

EF_mu 13_tight : Data vs MC Tag & Probe efficiency on data (Muons stream) and on MC (mc 10_7 Te. V. 106047. Pythia. Zmumu_no_filter. merge) Barrel Endcaps MC/Data = 1. 025 ± 0. 012 03/02/2011 MC/Data = 1. 007 ± 0. 006 A. Ventura - Fake probability in muon trigger chains 6

EF_mu 13_MG_tight : Data vs MC Tag & Probe efficiency on data (Muons stream) and on MC (mc 10_7 Te. V. 106047. Pythia. Zmumu_no_filter. merge) Barrel Endcaps MC/Data = 1. 010 ± 0. 012 03/02/2011 MC/Data = 0. 992 ± 0. 006 A. Ventura - Fake probability in muon trigger chains 7

L 2_mu 20 : Data vs MC Tag & Probe efficiency on data (Muons stream) and on MC (mc 10_7 Te. V. 106047. Pythia. Zmumu_no_filter. merge) Barrel Endcaps MC/Data = 1. 012 ± 0. 012 03/02/2011 MC/Data = 0. 992 ± 0. 006 A. Ventura - Fake probability in muon trigger chains 8

EF_mu 20 : Data vs MC Tag & Probe efficiency on data (Muons stream) and on MC (mc 10_7 Te. V. 106047. Pythia. Zmumu_no_filter. merge) Barrel Endcaps MC/Data = 1. 020 ± 0. 012 03/02/2011 MC/Data = 1. 005 ± 0. 006 A. Ventura - Fake probability in muon trigger chains 9

L 2_mu 40_MSonly : Data vs MC Tag & Probe efficiency on data (Muons stream) and on MC (mc 10_7 Te. V. 106047. Pythia. Zmumu_no_filter. merge) Barrel Endcaps MC/Data = 1. 064 ± 0. 029 03/02/2011 MC/Data = 1. 039 ± 0. 022 A. Ventura - Fake probability in muon trigger chains 10

EF_mu 40_MSonly : Data vs MC Tag & Probe efficiency on data (Muons stream) and on MC (mc 10_7 Te. V. 106047. Pythia. Zmumu_no_filter. merge) Barrel Endcaps MC/Data = 1. 120 ± 0. 024 03/02/2011 MC/Data = 1. 078 ± 0. 020 A. Ventura - Fake probability in muon trigger chains 11

Muon triggers of interest L 1 s ite m s em t i T HL mu 20 mu 13_MG_tight mu 13_tight Goal 03/02/2011 mu 40_MSonly is to check fraction of fakes in relevant muon triggers A. Ventura - Fake probability in muon trigger chains 12

Fake definition For each muon HLT signature, only the highest p. T firing feature (leading track) of the associated algorithm has been taken into account ◦ Suits single-muon p. T-based signatures, no double counts Signatures are tested looking for offline tracks from (Muid OR Staco) algorithms with p. T>2. 5 Ge. V Three definitions of fake for the leading track: a) no combined (or standalone in case of MSonly) offline tracks in DR < 0. 5 b) no combined (or standalone in case of MSonly) offline tracks at all (DR<999) c) no offline combined/standalone tracks at all (DR<999) (applicable only to L 2/EF combined signatures) 03/02/2011 A. Ventura - Fake probability in muon trigger chains 13

Fakes in EF_mu 13_tight Barrel Endcaps In all these plots vs p. T, the last bin concerns overflows (“infinite p. T”) Fakes are 0. 2% in barrel, 1% in endcap, 5% in transition regions 03/02/2011 A. Ventura - Fake probability in muon trigger chains 14

Fakes in EF_mu 13_MG_tight Barrel Endcaps 3% of tracks (>10% @>50 Ge. V) don’t match Muid or Staco CB 03/02/2011 A. Ventura - Fake probability in muon trigger chains 15

Fakes in L 2_mu 13_tight Barrel Endcaps Mu. Comb often triggers without Muid/Staco @high p. T and endcaps 03/02/2011 A. Ventura - Fake probability in muon trigger chains 16

Fakes in L 2_mu 20 Barrel Endcaps Similar Mu. Comb behaviour as L 2_mu 13_tight 03/02/2011 A. Ventura - Fake probability in muon trigger chains 17

Fakes in EF_mu 20 Barrel Endcaps Trig. Muon. EFCB has fakes around % (except in transition regions) 03/02/2011 A. Ventura - Fake probability in muon trigger chains 18

Fakes in L 2_mu 40_MSonly Barrel Endcaps Mu. Fast is almost fake-free in barrel, but 30% fakes in endcaps 03/02/2011 A. Ventura - Fake probability in muon trigger chains 19

Fakes in EF_mu 40_MSonly Barrel Endcaps “Infinite” p. T are often fakes; only 1/3 of L 2 fakes rejected in endcap 03/02/2011 A. Ventura - Fake probability in muon trigger chains 20

More on mu 40_MSonly Largest single muon HLT rates come from mu 40_MSonly L 2 has fakes mostly in endcaps EF has very little rejection power wrt L 2 (<7%) The apparently very large EF_mu 40_MSonly fake rate in the barrel looks probably due to bad Trig. Muon. EFSA estimate of h (muons presumably coming from endcaps) About 4% of Trig. Muon. EFSA fake tracks show |h|>2. 4 Trig. Muon. EFSA “infinite” p. T track in barrel are 90% fake 03/02/2011 A. Ventura - Fake probability in muon trigger chains 21

Fakes wrt different offline muons Using OR of muon offline references helps against calling “fake” some tracks in events with offline actually inefficient Example: Muons stream, Trig. Muon. EFCB EF_mu 13_tight ◦ Reference: FS = only Staco, FM = only Muid, FMS = Muid OR Staco Barrel Endcaps Definition of fake trigger could be improved involving other offline algorithms (reco/tag), or changing cuts… 03/02/2011 A. Ventura - Fake probability in muon trigger chains 22

Fakes on different streams (EF) Fake probabilities in Muons stream have been compared with Jet. Tau. Etmiss stream, using all EF_mu* items or orthogonal triggers (inclusive sample) ◦ Streams: Muons EF_mu*, Jet. Tau. Etmiss with EF orthogonal triggers Barrel Trig. Muon. EF 03/02/2011 Endcaps combined signatures look compatible A. Ventura - Fake probability in muon trigger chains 23

Fakes on different streams (L 2) L 2 signatures (both combined and MSonly) have different behaviour in endcaps: Muons stream looks more contaminated than Jet. Tau. Etmiss. Under investigation! ◦ Streams: Muons EF_mu*, Jet. Tau. Etmiss with EF orthogonal triggers Mu. Comb Similarly 03/02/2011 Mu. Fast EF_mu 40_MSonly: <1% fakes in Jet. Tau. Etmiss A. Ventura - Fake probability in muon trigger chains 24

Conclusions Good comparison in T&P efficiencies between data and MC ◦ Plateau ratio significantly ≠ 1 only in mu 40_MSonly Definition of “fake trigger” is arbitrary and can be depending on physics analysis ◦ Offline muons don’t always act as good reference: often fake trigger can be actually offline inefficiency ◦ Definition of DR wrt offline is not so relevant There are interesting hints on how to improve or refine hypotheses of some muon signatures Large mu 40_MSonly fake rate is still to be understood 03/02/2011 A. Ventura - Fake probability in muon trigger chains 25