1. TDAQ week Lisbon, October 2003 1 B-Physics Trigger Status Contents: Introduction: Di-muon triggers Hadronic Final States Muon-electron Final States Program of work & Status John Baines B-trigger Working Group

2. TDAQ week Lisbon, October 2003 2 B-physics Trigger: Introduction Flexible trigger strategy: di-muon trigger at L > 2 x 10 32 cm -2 s -1 Introducing other triggers at lower luminosities (lower lumi fills / later in beam-coast). Two strategies for these other triggers: Baseline: Require >1 LVL1 JET or EM RoI in addition to muon p T > ~8 GeV Perform track searches in RoI in SCT & Pixels (& TRT) Make selections based on reconstructed tracks, e.g. J/  (ee), B(  ), D s (  ) Reconstruction in RoI => saving in time, provided RoI multiplicity is not too high => Modest resources. Fallback (if RoI multiplicity too high or efficiency too low): Full-scan of SCT and pixels for events with muon p T > ~8 GeV To save resources, no TRT reconstruction => no J/  (ee) trigger Make selections based on reconstructed tracks, e.g. B(  ), D s (  ) => greater resources required, but better efficiency (lower p T threshold possible)

3. TDAQ week Lisbon, October 2003 3 single-muon di-muon all h h b b c c J/  @10 33 cm -2 s -1 Cross section (nb) In all cases muon trigger initiates B-trigger Cross-section for  from  an K falls more rapidly with p T than prompt  from b e.g. p T > 8 GeV : LVL1 rate 10 kHz for L = 10 33 LVL2 rate (muon & associated ID track) 2 kHz about 1/3 of rate due to b - >  Important not to raise muon threshold too high Instead, reduce rate by requiring other triggers TDR B-physics Di-muon triggers: Very effective trigger for e.g. B->J/  (  )K and B->  (X) LVL1 trigger efficient down to p T ~3 GeV (barrel) p T ~ 5 GeV (endcap) For 6 GeV threshold, LVL1 rate: 600Hz at L = 2 x 10 33 Mostly  from heavy flavours, plus some double- counted  in endcaps (later removed at LVL2) Muons confirmed at LVL2 using precision-  and ID EF selections based on mass and decay length cuts.

4. TDAQ week Lisbon, October 2003 4 Hadronic final states Baseline : Use LVL1 jet clusters to define RoI for ID track reconstruction at LVL2 (SCT and Pixels, optionally TRT) RoI ~ 10% of event = > promises substantial saving of resources provided RoI multiplicities low. Fallback: full-scan Studied with : Fast simulation including Longitudinal and transverse shower profiles, pulse history, digitisation and BCID => Multiplicity ~2 for E T > 5 GeV Full GEANT simulation, including noise, but without BCID. => gives higher multiplicities (expected as no BCID) => E T > 6 GeV gives efficiency of 80% for B s - > D s  events with B s p T > 16 GeV Final multiplicities expected to be closer to fast simulation results once BCID included Jet RoI Multiplicity E T (GeV) Event with muon p T > 6 GeV LVL2 selections for specific channels Based on ID tracks. Tracks refit at EF inside LVL2 RoI including secondary vertex fit. Expect final multiplicities to be closer to Fast simul. results Full Simul., No. BCID

5. TDAQ week Lisbon, October 2003 5 Muon-electron final states EM RoI Multiplicity E T (GeV) Used to select e.g.: B d -> J/  (ee)K s with opposite side muon tag B d ->J/  (  )K s with opposite sign e tag. LVL1 low E T EM clusters define RoI at LVL2 At LVL2, RoI confirmed as electrons in calorimeter Inside RoI tracks reconstructed in SCT pixels and TRT RoI small ~0.2 x ~0.2 => considerable saving in resources if RoI multiplicity not too high. Studied with fast simulation: Multiplicity ~1 for E T > 2 GeV Efficiency ~80% to find both e from J/y(ee) in events with e p T > 3 GeV Events with muon p T > 6 GeV Tracks refit at EF, including vertex fit and cuts applied to decay length and vertex fit quality Fast simulation results

6. TDAQ week Lisbon, October 2003 6 Rates (Hz) 2 x 10 33 10 33 LVL2EFLVL2EF B - >  (X) 200 small 100 small J/  (  ) 105 D s (  ) --609 B(  ) --203 J/  (ee) --102 Total 2001019020

7. TDAQ week Lisbon, October 2003 7 Work Programme - 1 TDR back-up note finished (Innsbruck + Alan Watson): ATL-COM-DAQ-2003-036. A study of the use of Low E T calorimeter RoI in the ATLAS B-Trigger ATLAS note on muon and di-muon cross-sections and rates in progress (Maria Smizanska) Measurement of LVL1 multiplicities and efficiencies based on full trigger simulation including trigger towers and BCID (Innsbruck + Alan Watson): Trigger code in place (Alan Watson) Currently waiting for help from LAr and Tile experts : Minor issues for Tile, Serious problems with LAr – noise too high, calibration stage missing time-scale unknown since experts busy working on test-beam s/w. => This is a major concern as it is currently impossible to get meaningful multiplicities Include LVL2 ID algorithms running in LVL1 RoI & generate NTuples (John, Nikos, Sergey) : Similar to e/  slice for J/  (e,e) trigger, but with B-physics datasets and initial layout Need to verify all components for initial layout Currently waiting for RegionSelector to be implemented for initial layout (Steve, Aline)  in the mean time, to enable analysis work to start, generate NTuples using 6.0.4 and data produced with DC1 layout

8. TDAQ week Lisbon, October 2003 8 Resource Issues – The Challenge Need to demonstrate a viable B-trigger within planned resources. Planning for intial system based on 8 GHz processor: LVL2 latency of 10ms @ 25kHz max. rate => 250 cpu (1000 for final system) EF latency of 1s @ 3.2kHz (?) => 3200 cpu Given current uncertainties in timings & overheads, how to compare resources? Compare B-trigger with EM25I trigger (50% of LVL1 rate) e/gamma trigger @ 2x10 33 B –trigger @ 10 33 8kHz EM25i x 2ms T2Calo = 16 cpu 10 kHz MU8I x 2ms MuFast = 20 cpu 2kHz em25i x t(ID FEX in 0.2x0.2 RoI) = 2T id 4 kHz mu8i x t(ID Fex in muon RoI) ~ 4T id Mult(jet) x 2kHz x t(ID FEX in 1x1 RoI) ~ 2M jet x 5? T id Mult(em) x 2kHz x t(ID FEX in 0.2x0.2 RoI) ~ 2M em x T id e/gamma = 16 + 2T id cpu B-trigger (RoI) = 20 + (4 + 10? M jet + 2 M em ) T id ~ 20 +23 T id B-trigger (full-scan) = 20 + 80? T id TestBed gives : T2Calo 5 ms on 2.6Ghz PC MuFast 8ms => both ~2ms for 8GHz PC Average no. SCT + Pixel RoB in 0.2x0.2 RoI : Total no. SCT & Pixel RoB :

9. TDAQ week Lisbon, October 2003 9 Work Programme Develop & compare strategies for J/  (e,e) trigger, e.g.: LVL1: muon RoI + 2 EM RoI LVL2: 2 calo e tags + vertex, decay length + J/  (e,e) mass LVL1: muon RoI + 1 jet RoI + 1 (2ndary) EM RoI LVL2: 1 calo e tag + J/  (e,e) mass (e + other track p T >5) + vertex + decay length LVL1 : muon RoI + 1 jet RoI LVL2 : 2 TRT e-tags, vertex, decay length + J/psi(e,e) mass LVL1: muon RoI + jet RoI LVL2: J/psi(e,e) mass (tracks pT>5) + vertex + decay length Measure Efficiencies for LVL2 RoI-based selections for J/  (e,e) and D s (  (KK),  ) Measure rates for B - >  X Measure speed-up obtained for xKalman/iPatRec running at EF inside LVL2 RoI c.f. full event reconstruction:  Requires xKalman and iPatRec running on data selected by RegionSelector. Make timing measurements & update resource estimates Include muon trigger simulation: Needs work on combined muon & ID tracks at LVL2 (muon trigger group) Need LVL1 di-muon trigger (muon trigger group) => ATLAS note on B-Trigger strategy planned for end 2003/early 2004

10. TDAQ week Lisbon, October 2003 10 Available Effort Some improvement to available effort: Aras Papadelis (Lund Diploma Student) Bill Scott (RAL ex. OPAL) Effort still very tight, especially people experienced with running Athena.  Problem : expert knowledge still required to use ATLAS software Continue to try to help with this by providing recipes on web pages. Will work on J/y trigger

11. TDAQ week Lisbon, October 2003 11 Summary Work going on on LVL1 RoI multiplicities & efficiencies Note on muon rates and cross-sections in progress Making progress with tools Major part of work still on hold awaiting missing infrastructure: RegionSelector for initial layout LAr calibration Aim for note on B-trigger strategy for end of year, but now looking very tight. Probably move to early 2004.  Making progress, but still a lot to do  Concerns over code stability during Autumn