1. Outline: HLT overview Objectives of muon+hadron Algorithm flow Selection parameters Monitoring Summary required Determination of parameters Antonio Pérez-Calero Hugo Ruiz

2. 31/05/2007Hugo Ruiz – Winter Meeting Santiago 20072 Single  L0  L0 di  Good hadrs Confirm Extra  s Di   + hadr Extra  s Recovr’d di   + hadrSingle  Generic B D* J/  Exclusive sels. ~ 1 KHz o(10 KHz) Confirm o(100 KHz) Alleys Global HLT reconstruction

3. Source of TOS for semileptonics: give ~ all semileptonics to excl/incl selection – Efficiency on B d  D*  : 31/05/2007Hugo Ruiz – Winter Meeting Santiago 20073 (Pessimistic)

4. Source of tagging-enriched TIS for any other channel: 4  10 9 unbiased B per year 31/05/2007Hugo Ruiz – Winter Meeting Santiago 20074 Better: count only events with other B in acceptance Even better: TIS efficiency on offl- selected evts And   tag eff

5. Exemple with the released version (  p T >1GeV, hadr p T >0.5GeV, both |IP| in [0.1,3]mm, DOCA 2mm) – 4.36 KHz, somewhere between alley and incl. selection For muonless samples of Hans’ preselection (Bs2Dspi, Bs2DsDs, Bd2Kpi, Bd2D0Kst, Bd2KstD2KSPiPi, Bs2PhiPhi, Bu2KD->KSPiPi, Bd2PiPiPi, Bu2Kee) – Events: 10,808 – L0: 55% – HltMuHad: 4.8% of those passing L0 – TOS: 2.25% of those passing L0 – TIS: 1.7 % of those passing L0 With extra tagging power Additional: nice muon id calibration sample? 31/05/2007Hugo Ruiz – Winter Meeting Santiago 20075

6. 31/05/2007Hugo Ruiz – Antonio Perez-Calero, HLT workshop at Amsterdam6 L0+T+VELO   p T,  IP VELO 3D hadron IP hadron p T VELO  forward Pointing, Dz VertexMaker: DOCA VertexUpgrader MuonHadron TrackFilter VertexFilter VertexUpgrader Reconstruction Legend: VertexMaker Sequencer 1 2 3 3 selection steps

7. Selection: 7 parameters (  2: alley & selection) –  : p T, Ipmin (Ipmax) – Hadron: p T, Ipmin (IPmax) – Vertices: DOCA, Pointing,  z Need still to study: – IP vs IPS – Can remove Ipmax? (yes from eff vs rate) Reconstruction: All parameters for: – Confirmed muon – VELO3D – VELO-forward 31/05/2007Hugo Ruiz – Winter Meeting Santiago 20077

8. What we need is what the standard algorithms will provide at the 3 selection steps: EVERYTHING This includes: – Input and output rate at each filtering level – # of input and output candidates at each filtering level Note that #filtering levels > #algorithms – Distributions of the variables in which we cut And of the ‘best’ of the candidates, to be able to extract the dependence of the rate on the variable Need: HLT-PVSS GUI that allows navigation in HLT alleys and algorithms to display rates, variable distributions… For shift team: rates + access to variable reference histos (with reference histos) 31/05/2007Hugo Ruiz – Winter Meeting Santiago 20078

9. 1.For TISTOSing: – Needed only summary from last HltVertexFilter Vertices that triggered Their two tracks – For TIS/TOSING based on summary to be a better approximation, we need the reco algos to give a chance to ALL good combinations – Not the best nth, for instance 31/05/2007Hugo Ruiz – Winter Meeting Santiago 20079

10. 2.To study alley performance (on offl-selected events) we need at each step: – Output candidates (LHCb id’s) – Cut values Extinguish “pretrigger and trigger” The info is made available from the standard algorithm summaries # objects in summaries: 31/05/2007Hugo Ruiz – Winter Meeting Santiago 200710 Filterscands Mu|IP| [0.1,3]mm 1,52 HltRecoVelo114 Had|IP| [0.1,3]mm DOCA<0.2mm 13,21 VELO=> Forward16,90 HadPT> 500MeV3,22 2nd DOCA cut3,76 Pointing<0.4 Dz>2mm 2,47

11. Aim: highest possible: 1.Semileptonic efficiency on alley 2.B purity on incl. selection for the allowed rate (muon alley will optimize muon confirmation efficiency) Rate at any set of parameters: from running HLT emulation on sample of random L0 triggered events Assume: – 1 st ~ 2fb -1 of data on disk – Then try to think about first months… 31/05/2007Hugo Ruiz – Winter Meeting Santiago 200711

12. 1.Alley: efficiency on TOS for semileptonics – Need to optimize it on TIS, running the HLT simulation – How many TIS semileptonics will have? B/S>0, but shouldn’t expect HLT to do better than offline! How many triggered on the other side? 15 M * 0.5 * 0.015 = 110 K events / year – Can use this events to optimize alley parameters Plot in slide 3 done with ~ 1 K events 31/05/2007Hugo Ruiz – Winter Meeting Santiago 200712 ChannelYield/2 fb -1 B/S B d 0  D * -  + 9 M0.4 B +  D 0 (*)  + 3.5 M0.6 B s  D s (*)  + 2 M0.6

13. 2.Incl. sel: efficiency on TIS for generic B – Take as starting point decays with highest rates: B +  J/  K + : 1.74 M events/year, B/S = 0.37 B d  J/  K * : 0.67 M events/year, B/S = 0.17 – >2M with kite good B/S – Require TOS: ~ 100% – Pass a typical muon+hadron: 2 M * 0.5 * 0.015 ~ 15 K events / year in the interesting region of an optimization plot (TIS efficiency vs rate). 31/05/2007Hugo Ruiz – Winter Meeting Santiago 200713

14. Assume 2  10 32, 10 7 sec/year: 1.Alley: 1.5 Hz of semileptonics (plot like in slide 2 in 1000 seconds, ~ 20’) 2.Incl. sel: – J/  K + /K*: 0.2 Hz – Need more. Can add more selections, in particular add semileptonics of point 1 to get a sample of ~ 2Hz Use them to evaluate trigger on the other side Here penalisation of factor ~ 0.5 * 0.015  need 30 hours to get same statistics 31/05/2007Hugo Ruiz – Winter Meeting Santiago 200714

15. If for these events we: – save the output of the HLT global reconstruction – send them to the monitoring farm We can obtain online Antonio’s plots: 31/05/2007Hugo Ruiz – Winter Meeting Santiago 200715 Muon from muon alley summaries Hadron from global reconstruction summaries Rate from monitoring of cut variables

16. At start-up lumi << 2  10 32 – Any combination of cuts will give << allowed rate – Just apply a set of minimum cuts to debug the alley! Under the carpet: – GEC/no GEC – Different luminosity setups 31/05/2007Hugo Ruiz – Winter Meeting Santiago 200716