Egamma Triggers

1. Egamma Triggers Alessandro Tricoli (CERN) On behalf of the egamma and trigger groups 20th October 2011 Level-1 Calorimeter Trigger Joint Meeting

2. Overview • Major E/Gamma trigger developments in 2011 and physics implications • Effort to maintain low thresholds at L1 and HLT and keep high performance • L1 adjustemnts • HLT selection tightening • Electron and Photon Trigger Performance • Outlook for 2012 • dead-material corrections at L1 • HLT isolation A. Tricoli

3. Overall Trigger conditions • New Trigger menu, Physics_pp_v3, already deployed since data-period L • covering lumi range ≥ 3x1033 cm-1s-1 • this year we will not reach 5x1033, but this menu is good preparation for 2012 • Run at 400 Hz total average EF output rate for the rest of this year p-p running • Total Rates within limits • L1: 50 kHz • L2: 4 kHz • EF: 500 Hz at peak 3x1033 cm-1s-1, average 400 Hz • Egamma stream: 110 Hz L1 L2 EF Egamma stream • Run 190256: • peak lumi 3.3e33 • 107 pb-1 collected A. Tricoli

4. Latest Primary Electron Triggers List of primary Electron triggers for remain of 2011 data-taking (rest of presentation will clarify their definitions and illustrate their performance) Primaries • Medium1 = medium++ offline • Loose1 = loose++ offline • See backup slides A. Tricoli

5. Latest Primary Photon Triggers List of primary Photon triggers for remain of 2011 data-taking Primaries A. Tricoli

6. 2011 rate constraints • During 2011 pp data-taking L1, L2 and EF rates had to be adjusted a few times to remain within budget • in winter menu was originally prepared for two lumi points: 5x1032 and 1x1033 • then adjusted for 3x1033 and planned for 5x1033 (never reached) • Limitations encountered at all trigger levels: • L1: raise thresholds • EM14->EM16; 2EM7->2EM10 • EM16->EM16VH; 2EM10->2EM10VH; add EM18VH as backup • low-pt triggers (i.e. J/Psi) still L1 limited • L2: a few times L2 selection of primary triggers had been tightened • low-pt tight triggers, e.g. J/Psi triggers • medium-pt tight triggers for tau studies, e.g. e15_tight • medium triggers, e.g. L2_e20_medium • medium1 triggers, e.g. L2_e22vh_medium1 • EF: raise of thresholds and tightening of identification selection • e20->e22, raise thresholds of combined triggers, e.g. e+met, e+mu • medium->medium1 (corresponding to medium++ offline selection, see backup slides) A. Tricoli

7. L1 threshold evolution Change of L1 thresholds since period L to reduce L1 rate without raising HLT thresholds • EM14 threshold removed to make space for other thresholds • Many more L1 items available from combinations of these 8 thresholds V = varied threshold, i.e. coarse dead-material corrections - effective ‘V’ threshold implements different thresholds in different eta regions with the coarse 0.4 granularity of the L1Calo cluster processor (e.g. EM16VH uses EM16,EM17,EM18) H = hadronic leakage requirement, i.e. absolute hadronic core requirement For detail description of ‘VH’ triggers refer to Monika Wielers’ talk on 26th July ATLAS Weekly https://indico.cern.ch/materialDisplay.py?contribId=1&materialId=slides&confId=119648 HLT triggers seeded of L1 ‘VH’ triggers have ‘vh’ added to the name, e.g. e22vh_medium1 • Under discussion the implementation of proper dead-material corrections next year A. Tricoli

8. L1 VH trigger definitions • For lumi ≥ 3x1033, L1 rate beyond safety margin, the choice was • raise L1 thresholds for trigger with high pT electrons and for H • apply different thresholds as a function of the coarse granularity of the cluster processor (=0.40.4) => V • Apply hadronic core isolation (= longitudial isolation in 22 towers behind EM cluster core) • very similar to hadronic leakage cuts applies in the HLT and offline • we wanted to avoid raise threshold (loss in physics acceptance) • we wanted to avoid EM isolation (unsafe due to pileup dependence) • dead-material corrections needed longer investigation • => Deployment of 2. and 3., i.e. VH, since August technical Stop • EM16VH • Hadcore ≤1 GeV (cts), ET thresh. cut between 16 and 18 GeV • rate reduction 45, rates at 3x1033 : EM16=18 kHz , EM16VH = 10 kHz • performance studied on Z events • EM10VH (to keep 2TAU8_EM10 within acceptable rate limits) • Hadcore ≤1 GeV (cts), ET thresh. cut between 10 and 11 GeV • rate reduction 40% on 2TAU8_EM10, factor 3 on 2EM10VH • performances studied on W and cross-checked on Z events A. Tricoli

9. L1 EM VH trigger performance Efficiencies of L1_EM16 and L1_EM16VH wrt offline medium++ electron identification by Z->ee tag-and-probe analysis • EM16VH turn-on curve steeper than EM16 and starting at slightly higher ET, • due to tighter thresholds in some eta bins, reach same values around ET=27GeV • details of turn-on profile under investigation eta region by eta region • only 0.1% efficiency loss of EM16VH wrt EM16 for Z→ee electrons with ET>25 GeV • Losses due to this hadronic leakage cut seen for electrons with ET>300-400GeV • EM30 [-> e45_medium1] does NOT apply hadronic leakage cut at L1 • High PT Electron analysis must use OR of triggers: • “e22vh_medium1 OR e45_medium1“ A. Tricoli

10. ‘T’ di-electron triggers • To reduce L1 rate with no increase of HLT thresholds on di-electron triggers • replace 2EM7 (18 kHz at 3e33) with 2EM10 (6.5 kHz at 3e33) then 2EM10VH (2 kHz at 3e33) • (VH introduced to cut rate of combined TAU+EM10 triggers) • bring L1 threshold closer to HLT threshold • Period A-J: 2EM7 seeding 2e12_medium • Period: K: 2EM10 seeding 2e12T_medium • Period L: 2EM10VH seeding 2e12Tvh_medium • (where ‘T’ indicates smaller gap between L1 and HLT thresholds) • Implications for physics analysis • for offline electron cut of ET>15 GeV the trigger is not fully efficient • => work on efficiency turn-on curve • Efficiency ratio 2e12T/2e12 shows • lower efficiency up to 20 GeV for 2e12T • at 15 GeV ~ 5% extra inefficiency for 2e12T Offline electron ET>15 GeV A. Tricoli

11. Evolution L2 and EF selections • L2 selection of lowest unprescaled single electron trigger tightened a few times this year • mostly with negligible impact on total trigger efficiency (sub-% effect) • In period L5, rate reduced by factor 2 in preparation for luminosities ≥ 4 x 1033 cm-2 s-1 • tighter shower shape cuts and tighter Dh cut between cluster and track • L2_e22vh_medium1 rate from 800 Hz to 400 Hz at 3e33, and efficiency wrt offline medium++ decreased by 1% • Evolution of lowest unprescaled single electron trigger • (to accommodate rate within allocated bandwidth) • periods A-J: e20_medium • (projected rate at 3x1033 ~170 Hz) • period K: e22_medium • (projected rate at 3x1033 ~130 Hz) • period L: e22vh_medium1 • (rate at 3x1033 ~60 Hz) • – medium1 same identification selection as medium++ offline (see backup slide) • Implication on physics analysis: • slightly lower efficiency • offline electron that fired trigger must be selected with medium++ identification A. Tricoli

12. e22vh_medium1 performance efficiency Z→ee tag-and-probe Z→ee tag-and-probe Z→ee tag-and-probe • efficiencies of e20_medium and e22_medium comparable on plateau • drop by ~2% in transition medium→medium1 • most losses due to tracking at high h • detail investigation on-going • e22vh_medium1 efficiency in period L5, measured by Z→ee tag-and-probe for ET> 25 GeV L1: 99.83±0.02 L2: 97.61±0.06 EF: 95.51±0.09 A. Tricoli

13. e22vh_medium1 performance pileup dependence Integrated efficiency of e22(vh)_medium1 wrt offline medium++ for electron ET>25 GeV in period F-L, as function of no. of primary vertices • N.B.: trigger efficiency measured relative • to same offline selection • pileup dependence may largely cancel • here shown additional pileup dependence of trigger relative to offline (due to resolution effects) • L1 shows no pileup dependence • L2 and EF efficiency show pileup dependence • on-going investigation A. Tricoli

14. Photon trigger performance • Single photon triggers in 2011: • unprescaled: g60_loose-> g80_loose • prescaled: g20_loose, g40_loose, then g60_loose • di-photon trigger: • unprescaled: 2g20_loose – to be kept unprescaled for the rest of the year • EM14 threshold removed from period L => (2)g20 triggers re-seeded of (2)EM12 • photon trigger performance very stable throughout the year • g20_loose efficiency close to 100% • efficiency measured by bootstraping from low threshold L1-pass-through triggers EF_g20_loose efficiency Bootstrapped from L1_EM12 Wrt offline tight photons Periods L2-L3 • L1 Pass-through triggers (no HLT selection): • EM3 - 1 Hz • EM12 - 1 Hz • 2EM12 - 1 Hz (to be disabled) • EM30 - 1 Hz ET [GeV] A. Tricoli

15. Jpsi triggers Electron efficiency calculated from J/ψ ll, W lν, Z ll data samples: use J/Psi and W triggers (see following talk by Lydia) • J/Psi triggers • effort to maintain J/Psi tag-and-probe triggers at 3x1033 luminosity. • crucial to understand low-pT electrons for e.g. H → 4e • J/Psi -> ee events are also important for the understanding of the EM calorimeter performance (extraction of resolution, intercalibration, etc) • Tag electron is tight; no identification selection on Probe electron leg (only ET cut) • Invariant mass cut 1-5 GeV at HLT to improve purity • Jpsi triggers are all prescaled • => statistics is a concern • Limitation from L1 rate • =>L1 rate increased progressively • during a fill when lumi and total L1 • rate drop Example where L1 topological triggers would be very useful (e.g. inv mass cut at L1 as done at HLT) A. Tricoli

16. W tag-and-probe triggers Electron efficiency calculated from J/ψ ll, W lν, Z ll data samples: use J/Psi and W triggers (see following talk by Lydia) • W tag-and-probe triggers • Tag the event via missing ET significance XS~MET/sqrt(SumET) • apply minimal requirements Probe electron: only ET threshold and basic track quality cuts • to further reduce rate and allow trigger to run unprescaled, MET isolation applied • Df(MET, jet) > 0.7 or > 2.0, with jet pt>10 GeV • Run 190256: • peak lumi 3.3e33 • 107 pb-1 collected • Rates show clear dependence on pileup • => rate increase during a fill as pileup decreases • missing ET significance inefficient at high pileup • bias on offline efficiency measurement in different pileup conditions under investigation A. Tricoli

17. Prospects for dead-material corr. Preparation for 2012 data-taking • dead-material correction could replace ‘V’ thresholds • provide sharper turn-on curve • efficiency vs eta is flatter • efficiency plateau reached at lower ET • results with Zee tag-and-probe so far are encouraging • need to adjust thresholds seeding HLT chains to control rates • list of L1_EM threshold would need to be revisited • need to cross check results on lower-ET physics signals, e.g. J/Psi • other trigger signature groups need to validate the results obtained within egamma A. Tricoli

18. HLT Isolation Preparation for 2012 data-taking • EF_e22vh_medium1 projected rate at 5x1033 is ~96 Hz • rate reduction by factor 3 by tightening medium1 selection and add EF isolation • Egamma proposal for EF isolation – two options • both are OR between one isolated and one higher threshold non-isolated trigger • opt. 1: raise threshold from e22 to e25, tighter identification selection, loose relative track isolation OR non-isolated higher threshold trigger • opt. 2: keep e22 threshold, tighter identification selection, tight relative track isolation and loose relative calo isolation OR non-isolated higher threshold trigger • (see backup slide) Studies on Z->ee show good performance of track and calo isolation at EF wrt offline variables • some analyses want to and can avoid trigger isolation by using combined triggers • defining combined triggers to reduce rate and avoid isolation on electron trigger • thriving discussion on what triggers to implement compatibly with rate budget • electron+jets, electron+MET, di-electron, electron+muon full scan • possibly more… • More supporting triggers to be introduced A. Tricoli

19. Conclusions • Excellent performance of L1 EM thresholds in 2011 (thanks!) • Continuing effort to keep egamma triggers with the lowest possible threshold and highest performance with ever increasing LHC luminosity • adjustments at L1 (V, H and T triggers) • tightening of selection at HLT • Continuing studies of trigger performance • Preparatory work for 2012 • positive studies of dead-material corrections • proposal for electron trigger isolation to keep single electron trigger at lowest possible threshold • egamma workshop 24-28 October 2011, includes triggers session • summarise 2011 operations and performance, and plan 2012 • https://indico.cern.ch/conferenceDisplay.py?confId=148154 A. Tricoli

20. Backup A. Tricoli

21. Medium++ and loose++ • Medium1 at trigger is equivalent to medium++ offline • Loose1 at trigger is equivalent to loose++ offline • medium++ • medium selection requirements with the addition of: • tighter shower shapes for |h| > 2.01 • tighter Dh track-cluster matching ( | Dh |<0.005) • stricter bLayer and Pixel hit requirements • bL + bLOutlier > 0 for |h| < 2.01 and nPix+nPixOutliers > 0; • nPix + nPixOutliers > 1 for |h| > 2.01 • loose TRT HT Fraction cuts • loose++ • looser shower shape cuts than loose • additional tracking cuts as in medium • nPix+nPixOutliers >=1 and nSi+nSiOutliers >= 7 • loose Dh track-cluster matching ( | Dh | < 0.015) For more details see Egamma Twiki page https://twiki.cern.ch/twiki/bin/view/AtlasProtected/TechnicalitiesForMedium1 A. Tricoli

22. e12_medium performance Efficiencies of combined triggers, e.g. 2e12(Tvh)_medium, obtained from efficiency of single e12_medium trigger • 2011 Data: 1.5 fb-1, period B-J • MC10b: 5 fb-1, 106046 Zee • Zee tag-and-probe analysis • Tag electron: tight and isolated electron that trigger e20_medium, exclude crack • Probe electron: ET > 12 GeV, |η| < 2.47 data mc data mc e12_medium wrt tight e12_medium wrt tight Scale factors for e12_medium wrt offline tight(medium) electrons available https://indico.cern.ch/getFile.py/access?contribId=1&resId=0&materialId=slides&confId=116220 A. Tricoli

23. Current Supporting Triggers • Supporting Triggers for electron background measurements: • e11_etcut – 0.5 Hz • g20_etcut – 1.5 Hz • e22vh_loose – 1.5 Hz • e22vh_loose1 – 1.0 Hz (loose1 = loose++ offline selection) • e22vh_looseTrk – 1.0 Hz (looseTrk=loose + tack quality cuts) • e22vh_loose_3j20_a4tc_EFFS – 0.5 Hz • e22vh_loose_4j15_a4tc_EFFS – 1.0 Hz • Jpsi and W tag-and-probe triggers (see next slides) • Supporting triggers for tracking studies: • e22vh_medium_SiTrk, e22vh_medium_TRT, e22vh_medium_IDTrkNoCut – 0.5 Hz • L1 Pass-through triggers (no HLT selection): • EM3 - 1 Hz • EM12 - 1 Hz • 2EM12 - 1 Hz (to be disabled) • EM30 - 1 Hz A. Tricoli

24. HLT Isolation Preparation for 2012 data-taking • EF_e22vh_medium1 projected rate at 5x1033 is ~96 Hz • rate reduction by factor 3 by tightening medium1 selection and add EF isolation • Plan to run isolated electron triggers soon, for validation studies only, but NOT for event selection until next year 1) pt >25, bLayer cut added also in |h|>2, ptCone20/pt < 0.100 OR’ed with pt >30, bLayer added also in |h|>2 2) pt>22, bLayer also in |h|>2, ptCone20/pt > 0.050, etCone20/et < 0.175, OR’ed with pt >30, bLayer added also in |h|>2 e25i5vh_medium2 [30 Hz at 5e33] e30vh_medium2 [+5 Hz unique rate] • Egamma proposal for EF isolation – two options • both are OR between one isolated and one higher threshold non-isolated trigger • opt. 1: raise threshold from e22 to e25, tighter identification selection, loose relative track isolation OR non-isolated higher threshold trigger • opt. 2: keep e22 threshold, tighter identification selection, tight relative track isolation and loose relative calo isolation OR non-isolated higher threshold trigger e22i6vh_medium2 [35 Hz at 5e33] e30vh_medium2 [+9 Hz unique rate] For more details see, egamma talk at Trigger General Meeting: https://indico.cern.ch/getFile.py/access?contribId=4&resId=0&materialId=slides&confId=129126 Studies on Z->ee show good performance of track and calo isolation at EF wrt offline variables A. Tricoli