1. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 1 Tagging and Offline selection versus multiplicities What are the “best” multiplicity variables to be used in the trigger ? Take muon tagging as a study case Raise a few questions, but give no definitive answers due to lack of statistics (waiting for DC04)

2. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 2 Bs->KK MC truth study: 26790 offline-selected 2145 direct-from-B MC tag muons 283 wrongly charged 1365 hit VELO+T+MS* 1258 matched to Particle 152 are wrongly charged (oscillation) First conclusion: the good MC tag muons are found and reconstructed quite efficiently (~92%). Difference 2145-1365=780: ~300/480 are outside/inside VELO acceptance and about 200 of the 480 would have more than 3 TT layers hit (of which ~120 have also enough T hits). Typically low- momentum, still … seems hopeless to do any PID with RICH only. Anyway… In a dilution-free world we should get (a):  tag = 5.1 % w = 11 %  eff = 3.1 % Compare to “official” tagging results (b):  tag = 6.3 % w = 37.2 %  eff = 0.4 % Naïve: Assume we manage to select only a fraction f of the above 1433 ones and we add N rnd “randomly charged” tag muons ( N’ tag = N tag f + N rnd and N’ wrong = N wrong f + 0.5 N rnd ). Then, we need N rnd = 1204 and f = 0.43 in order to go from (a) to (b). Not exactly 10% BR, because of cut  <0.5rad for storing a MCParticle… * #velo sensors > 5 #T-layers > 5 #muon-stations > 2 How many good tag muons ?   P<5GeV/c 2

3. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 3 Distributions for various Muons Split into 6 categories of Muons: Muons matched to a MC non-muon (mostly pions) Muons matched to an MC muon coming from u/d mesons (mostly pions) bottomed hadrons charmed mesons kaons other sources “Efficiency”  DLL  cut = Particles with Muon System info and with DLL  > -9 “Wrong-tag fraction” w Main contaminants, random charge (w=0.5) Less frequent but wrong charge! (w=0.75) BR: b  - X  10.7 % b  c  + X  8.3 % s + W+W+ b b B (signal) PV B c - W-W- b

4. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 4 P and P T Distributions for various Muons Note:  is not quite the  -tag algorithm efficiency: here, all  ’s are taken, even if there are n>1  ’s per event. Clearly, the P T cut is the main handles to distinguish the “bad” (in particular the “charm”) muons from the “good” muons. P cut P T cut MeV/c 2 “nominal” cuts

5. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 5 Puips = Min(ips) for all Piled-Up Primvertices Watch out for the overflow bin (cases with no piled-up primvtx) After removing the low- momentum particles, one clearly sees the tag-muon candidates coming from the wrong (pile-up) primary vtx: 2 peaks are well separated! It seems there is still quite some room to tighten the cut (good for higher lumi?) Hint that (muon) tagging algo is not much affected by presence of piled-up vertices P > 5 GeV/c 2 P T > 1.2 GeV/c 2 P > 5 GeV/c 2 P T > 1.2 GeV/c 2 ln(puips) cut No cut

6. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 6 IPS cut P > 5 GeV/c 2 P T > 1.2 GeV/c 2 Puips > 3.7 P > 5 GeV/c 2 P T > 1.2 GeV/c 2 Puips > 3.7 No. Hence, should allow reducing the contribution (31%) of misIded , decaying  and K, by a factor  3. Statistics too low to give the gain in tagging power Not tried using the fwd displacement of muons w.r.t. to PV Cut usable in L1 Note: w in-/de-creases with IPS cut. Correlation with oscillation (via proper time) ? Cut also on IPS to signal PrimVtx ? Pions come from primvtx, not the good muons Currently not used in official tagging Is the discriminating power correlated with low-momentum ? (hence, reduced discrimina- tion after momentum cut ?) IPS cut No cuts

7. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 7 Tagging vs multiplicities and number of primary vertices So far, looked only at “global” multiplicity variables (PU multiplicity, SPD multiplicity, VELO cluster multiplicity) Now, try to disentangle whether tagging depends on single vertex multiplicity, or on number of vertices, or whatever … All following plots done with same 27k Bs=>KK offline selected events (no trigger).

8. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 8 Tagging vs Total Multiplicity N tot Tagging  fficiency strongly increases with multiplicity Wrong-tag fraction as well Tagging Power: best at low multiplicities But is it due to the correlation of N tot and N prim ? Any N prim Tagging flags from OD’s event lists N tot = parts = the number of DV Particles (~ VTT+Long tracks in my case) N prim = prim = the number of reconstructed primary vertices

9. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 9 Tagging vs Number of Rec. Prim. Vtx N prim Tagging  fficiency seems NOT to depend much on number of prim vertices (except perhaps the same- side kaon tag ?) Wrong-tag fraction and tagging Power … ? Any N tot

10. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 10 vxtmul(i) = the number of velo tracks used to form the primary vtx i Is the tagging mostly affected by multiplicity of the B primvtx (i =1) rather than by the global multiplicity and/or the number of primary vertices ? 15<vtxmul(1)<35 vtxmul(1)>50 Tagging vs N prim for Given Vtxmul(1)

11. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 11 Tagging vs Multiplicity for selected N prim Use the number of long tracks from the signal vertex as a measure of multiplicity, in order to avoid N prim % N tot correlations Differences between two plots are weak (?) Dependence on multiplicity is strong N prim >1 N prim =1

12. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 12 Versus multiplicity, including offline selection efficiency Here Tag = combined tag Vtx1 = number of long tracks of highest multiplicity primary vtx Tot power = offsel efficiency x tag power N prim =1 N prim >1

13. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 13 … now versus number of primary vertices 8  Vtx1 < 16 24  Vtx1< 32 Here Tag = combined tag Vtx1 = number of long tracks of highest multiplicity primary vtx Tot power = offsel efficiency x tag power

14. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 14 PU detection vs vtx multiplicity measurement The tagging efficiency seems to be mostly sensitive to the signal** vertex multiplicity The offline selection seems somewhat sensitive to the number of primary vertices Altogether (offsel+tagging), the signal vertex multiplicity seems to be the most critical quantity If that is confirmed, should we not use the PU detector differently : look at peak1 size rather than peak2 size? Or, at least, combine the various quantities Then, should we not rather put the PU forward* ? What !? Are you nuts ? * In fact, even if we cut on the 2 nd peak, how big is the chance that a fwd PU sees the signal B vertices as a 2 nd vtx ? ** Note: when N prim >1 the signal vertex is the highest multiplicity vertex in about 2 out of 3 events.

15. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 15 Correlation between fwd and bwd tracks Backward VELO tracks vs forward VELO tracks (from signal prim vertex) Backward VELO tracks vs forward LONG tracks (from signal prim vertex)

16. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 16 Back to Muon Categories misId   ->  decay good charm K->  decay other  There seems to be also a dependence, on number of primvertices, but remember the correlation to multiplicity… Apply nominal tagging cuts (1.2 GeV/c 2 P T, 5 GeV/c 2 P, 3.7 Puips) Plot the muon category fractions vs multiplicity variables (not integrated plots!) The lower the number of physics Particles, the purer the tagging muon sample: The higher the number of physics Particles, the more pion-related muons we get…

17. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 17 DV Particles (Long + VTT) What is the Best Track Multiplicity Variable ? Velo Tracks from B primvtx Backward tracks from B primvtx Purity depends coarsely (?) on backward tracks from B primvtx (PU detector !) Long tracks from B primvtx Almost the same for long tracks from B primvtx as for Particles But what will happen at higher Lumi ?

18. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 18 Summary Although the statistical accuracy does not yet allow solid conclusions, it seems that offline+tagging is more sensitive to the signal vertex multiplicity than to the number of primary vertices (or the total multiplicity) L0 PU: does one gain by cutting on peak1 size rather than peak2 size (or combining the two) Let’s wait for DC04 before jumping to conclusions… Just one more slide

19. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 19 Tag Muon Availability vs PrimVtx Multiplicity Try to answer for tagging muons. Always using 27k Bs=>KK Plot tagging muon “availability” as a function of number of Particles per event (with a single reconstructed primary vertex) Availability = fraction of “direct- from-B” MC muons which leave info in at least 6 VELO sensors, 6 T layers and 3 Muon stations. Can’t see a drastic drop at low Multiplicities. Below ~20 Particles, there might be a slight loss (but there aren’t many events there !) Question raised in last Monday’s meeting: can we loose (good) tagging particles if we cut too hard on multiplicity ? avge Particles per event

20. LHCb Trigger Meeting – Februaryr 9, 2004, CERNMassimiliano Ferro-Luzzi 20 No striking difference between N prim =1 and N prim > 1, except for the same-side kaon tag: indeed, one expects a stronger increase of  for N prim =1 than N prim >1 (in the latter case the multiplicity is shared by several vertices) Is it due to the multiplicity of the signal vertex ? Tagging vs N tot for given N prim N prim >1 N prim =1