1. Trigger study on photon slice Yuan Li Feb 27 th, 2009 LPNHE ATLAS group meeting

2. Outline Introduction Trigger efficiency study in H->  and comparison with offline Photon trigger efficiency measuring methods summary

3. Brief Introduction of ATLAS trigger structure L1 Build RoI L2 Build TE Based on L1 RoIs EF Calculate more information Based on L2 TEs Use coarse- granularity information of calorimeter and muon detector Full detector information(in cluding the ID), refined granularity Full detector information, use as much as possible offline reconstruction algorithms and tools 40M Hz 75k Hz 2k Hz 200 Hz

4. Trigger efficiency study in H->  and comparison with offline Motivation: Repeat the study on trigger effect in H->  analysis done for the CSC exercise Samples & software : mc08 (10TeV, lumi 10 31, H->  &  +jet) athena 14.2.23, HiggsAnalysisUtils Offline selection: cut-based photonID (optimized for H->  analysis, very similar to isEM) |eta|<2.37, excl. [1.37, 1.52], track isolation (track pt sum in ΔR<0.3 + additional cut in ΔR<0.1) kinematic cuts(Pt  1 >40 GeV, Pt  2 >25 GeV) Apply Trigger 2g17i after offline selection: same trigger Item as in CSC: L1_2EM13I, L2_2g17i_L33, EF_2g17i_L33

5. Trigger efficiency w.r.t offline selection L1L2EF CSC results [%] Signal96.3±0.395±0.493.6±0.4 My results [%] Signal99.0±0.298.9±0.294.7±0.5 Background90.1±0.689.3±0.680.2±0.8 Signal (H->  ): 4k events Background (  +jet): ~3M events CSC data: 14 TeV and 10 33 (small effect) release 12.0.6 L1 and EF still reject a lot of events (especially background) after offline cuts mc08 data: 10 TeV and 10 31 release 14.2.20.3 L1 calorimetric isolation is looser than rel12

6. Comparison between L1 and offline: calorimetric isolation L1 effect due to calorimetric isolation already noticed at CSC time. Try to understand that effect by checking calorimetric isolation offline using etconeNoise30 ( 3σ above total noise for EM + HAD cells, 0.1<ΔR<0.3 ) see Caroline Collard’s talkCaroline Collard’s talk This variable is only available in ESD samples Offline+L1Offline+etconeNoi se30 H->γγ [%]99.0±0.2 γ+jet[%]90.6±0.582.5±0.7 Further study needed with pile-up samples to conclude on the usefulness of isolation. Calorimetric isolation efficiency w.r.t offline. 6

7. Differences between EF and offline photon selection MC samples were produced with the old Trigger selection: Variables used not strictly the same: e233/e277 in EF and e233/e237 in offline Et in EF and Pt in offline different η definition for fiducial cut Bug in the EF hadronic transverse energy in 1st layer (ethad1) Cuts were different and sometimes tighter at EF than offline All those problems are known and (being) fixed 7

8. Relative difference between EF and offline variables Layer 2 Layer 1 e237/e277 weta2 f1wtots1 weta1 fracs1    8

9. 2g20 study g20 is one of the primary triggers for first data 2g20 can be used to measure g20 efficiency (“tag&probe” method) Differences w.r.t. old 2g17i: – No isolation being applied at L1 – Use isEM to perform the EF selection (isEM&0xFF==0) EM Layer 2 variables only: e237/e277, e233/e237, weta2 hadronic leakage (ethad1/Et) Ask for at least 2 EF objects to pass the isEM selection 9

10. 2g20 efficiency L1L2EF Signal[%]99.9±0.199.8±0.199.4±0.2 Background[%]99.8±0.1 97.8±0.3 Much higher L1 efficiency using 2g20 (no isolation). High EF efficiency as expected with looser ID cuts Warning : EF_2g20 trigger decision was made by myself, not official code. But consistent with Martin Tripiana’s checks on corrected EF code for release 15 10

11. Photon trigger efficiency measurement “tag & probe ” method to measure photon trigger efficiency Bootstrap method collect the photon sample with photons passing L1_EM13, then measure the efficiency of matching g20 trigger object g20   ag  probe   ag required to match with g20(no cuts on this photon),  probe required to pass tight photonID cut and track isolation. No pt cut applied.

12. Photon trigger efficiency

14. summary Study trigger efficiency of 2g17i w.r.t offline – discrepancies between EF and Offline codes understood; – EF and Offline variables can be different especially those related to calorimeter layer 1; Most of the differences disappear in (present) 2g20 Measurement of g20 trigger efficiency – “Tag&Probe” method – Bootstrap method

15. backup

16. L1 calorimeter selection Based on a window of 4×4 trigger towers, each tower is Region of Interest(RoI): a local Et maximum in the 2×2 trigger towers. Egamma candidate: 1, RoI is a local Et maximum 2, pass the energy theshold Isolation requirements: 1, EM Et of 12 towers around the RoI below a threshold 2, Hadronic Et of 12 towers around the RoI below a threshold 3, Hadronic Et of 4 towers behind the RoI below a threshold

17. L2 calorimeter selection Scan the EM 2nd layer in a region of 4×4 trigger towers around the seed position. The cell with the highest energy is found and used as a base to build a cluster of 3×7 cells Selection variables: 1, EM transverse energy Et in the region of 2, Hadronic transverse energy 3, The 2nd layer: 4, The 1st layer: For electron, three more variables:

18. EF selection ET (EM) f1 ET Had e237/e277 weta2 e233/e277 e2tsts1/emisn1 wtots1 fracs1 weta1 At EF trigger level, the egamma selection is offline-like analysis but with a looser selection. Calorimeter information is used to select events containing the high Et EM shower. Selection variables: