1 / 44

EbE Vertexing for Mixing

EbE Vertexing for Mixing. Alex For the LBLB group. Status. Increased sample’s statistics Full ~350 pb -1 D 0 , D + , K + , K*, ’ Primary Vertex : SF robustly sitting around 1.38 Dependencies (Z,Pt,Si hits,Ntracks) Effect of hourglass

min
Télécharger la présentation

EbE Vertexing for Mixing

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. EbE Vertexing for Mixing Alex For the LBLB group

  2. Status • Increased sample’s statistics • Full ~350 pb-1 • D0, D+, K+, K*, ’ • Primary Vertex: • SF robustly sitting around 1.38 • Dependencies (Z,Pt,Si hits,Ntracks) • Effect of hourglass • G3X (no time to show today, but does not seem relevant with current statistics) • Systematics • Comparison of pulls and extraction of a common value: • D0 vs Lxy • Secondary Vertex: • Dependencies (Z,,, Pt,Lxy,ct, ,R,Isolation,Si hits) • Extraction of a common scale factor with systematics?

  3. Primary Vertex

  4. The tools • Prompt peak • V-truth • V1-V2 • d0/ B Lxy d0

  5. Scale Factor from V1-V2 • Fit two independent subsets of ‘primary’ [I.e. non-B] tracks • Measure (x1,y1,z1) and (x2,y2,z2) • Obtain / for x, y and z • Fit core with single gaussian (central value) • Repeat fit with two • gaussians (‘syst.’) • Still using 1.38 • For what follows

  6. Is the PVSF ‘universal’?

  7. ’ as a probe the PV scale factor • Prompt production • Lxy wrt EbE is 0! 1.310.02 1.010.04 (~86% of the distribution) NB: the other studies indicate a SF of ~1.38 The number I get here with the systematic uncertainty is consistent. Systematic is wide because of the presence of displaced ’

  8. Bottomline… • PV Scale factor shows no strong dependence on variables probed • Any other variable you want to see? • We could be more accurate but remember that the statistics is limited! • Focus on assessing systematics • Inter-sample variation • Change fit model (just like the overall fit) • Will try to improve a little more on statistics (K, D), but mostly aiming at perfecting systematics

  9. We can use the B I.P. pulls as cross check of the Lxy resolution… Do Impact Parameters Give a consistent picture? • Apply 1.38x and beamline constraint, check what happens: • For instance with D0: EbE: 1.351.15 • Unexpected? Not quite: see plot on the left! • There are two additional sources that enter in both cases: • Hourglass parameterization (including time-dependancy: see Aart’s talk) • A secondary vertex scale factor

  10. Relevance of the SV scale factor • d0/Lxy uncertainty is a combination of: • PV covariance • Beamline covariance • SV covariance • So far only PV was discussed! • PV scale factor is not the full story: when you bring down PV, the importance of SV increases • Need to get the SV scale factor right! • REM: even if PVSF=SVSF, we cannot use one common Lxy SF (the beamline covariance enters too into the expression!) 1.38x on PV 1.38x on SV This scale factor is just an initial guess, based on what we see on PV

  11. Secondary Vertex

  12. Scale factor from B decays • Example: BK+ • Fit  to a single vertex • “point”  back to K • Measure Lxy wrt B vertex • Pull is a proxy for a “seconday vertex” pull!   “Two” K B “One” Primary Vertex

  13. Samples and Topologies used: “3-1” • BK+ (1:K2: ) • BK* (1:K2: ) • D+K (1:2:K ) • ’ (1:2:) • (1:2: ) 2 1 “2-2” 2 1

  14. Bottomline for SV • Pull grows as a function of lifetime!@#^$! • Hidden dependencies! • Detector acceptance? • Kinematics? • Multiplicity? (no: K*) • Figure out which distributions are different • Check dependency! …results in the next pages

  15. Pt of ‘one’ Z Lxy Pull Distributions Ct of ‘one’ one’s Lxy  Of two Isol. (0.7) Pt of two Pt one’s tracks #L00 tracks  Of two R  • Detector acceptance (z) pretty similar • No clear difference in Si properties • Kinematics differs ( R Pt) • However… #stereo tracks Pt at <0.15 Pt at >0.15

  16. Pt of ‘one’ Pulls vs variables in prev. page Z Lxy Pull Ct of ‘one’ one’s Lxy  Of two Isol. (0.7) Pt of two  Of two Pt one’s tracks #L00 tracks R  • Comparing ’ to average of other samples in each bin • Everything excluded except ct • Why? I can think of possible reasons, but in terms of bugs mostly! WORK IN PROGRESS #stereo tracks Pt at <0.15 Pt at >0.15

  17. Ct and Lxy Ct(one) distribution Lxy(two) pulls vs ct(B) D+ ’ K* ’ !!! Lxy distribution Lxy(two) pulls vs Lxy(B) ’ !!! The problem does not show up with prompt objects!

  18. Ct(one) distribution D+ Montecarlo vs data Lxy(two) pulls vs ct(B) Data MC Lxy(one) distribution Lxy(two) pulls vs Lxy(B) The problem does not show up in montecarlo!

  19. ’ data vs D+ MC Lxy(two) pulls vs ct(B) Lxy(two) pulls vs Lxy(B) D+ MC ’ data • They are much more similar! • The ‘bug’ affects non-prompt data only!!!

  20. Lxy(Lxy) and pulls for’ Lxy mean (Lxy) mean Lxy/(Lxy) mean Lxy(D+) Lxy(D+) Lxy(D+) Lxy width (Lxy) width Lxy/(Lxy) width Lxy(D+) Lxy(D+) Lxy(D+)

  21. Same plots in D+ data: Is it in (Lxy) or in Lxy? Lxy mean (Lxy) mean Lxy/(Lxy) mean Lxy(D+) Lxy(D+) Lxy(D+) Lxy width (Lxy) width Lxy/(Lxy) width Lxy(D+) Lxy(D+) Lxy(D+)

  22. Same plots for MC D+ Lxy mean (Lxy) mean Lxy/(Lxy) mean Lxy(D+) Lxy(D+) Lxy(D+) Lxy width (Lxy) width Lxy/(Lxy) width Lxy(D+) Lxy(D+) Lxy(D+)

  23. Same plots for D+ without L00 Lxy mean (Lxy) mean Lxy/(Lxy) mean Lxy(D+) Lxy(D+) Lxy(D+) Lxy width (Lxy) width Lxy/(Lxy) width Lxy(D+) Lxy(D+) Lxy(D+)

  24. Secondary Vertex Conclusions, so far • Surprising Dependency on ct • Problem shows up ‘only’ in long lived signal in data • it’s a pity that’s what we want to use for our analyses ;) • Semileptonic lifetime? (biases are present as well!) • Working on finding the cause • Montecarlo: It works! • Swimming of track’s covariance to the vertex (CTVMFT does not account for that): no effect • Investigate other variables in data (Impact parameter?) • Probe other samples (D0?) • The problem I see is consistent with Aart’s findings on the impact parameter scale factor as a function of (K-) • This is the last pending big issue at the moment

  25. Moving along the plans for improvements! • Understand beamline parameterization: • Is it modeled correctly • Is it measured correctly  Include our best knowledge of it! • Are secondary vertex pulls ok? • Check with montecarlo truth • Use n-prong vertices (J/K, K+/0, K+/0) • Investigate dependencies (Pt, z,multiplicity, )with full statistics

  26. Plan • PV shows a very consistent picture • Finish up PV studies (~days): • re-run some ntuples to get full statistics in all cases • Including K, D • Use also BK background to get a source of studies for prompt Lxy pulls? • SV riddle to be solved! • I am working full steam on this. • One more weeks according to schedule, to straighten everything out, document and insert in the blessing pipeline.

  27. Plots a la 7500

  28. Primary Vertex Pulls for: • All B signals (left) • ’ (below) R Isol(R<0.7) Z R Z Isol(R<0.7)  Pt Just no statistics!  Pt Non-statistical fluctuations dominated by fit model!

  29. Z R Isol(R<0.7) • Secondary Vertex Pulls for: • All B signals (left) • ’ (below) Pt  Isol(R<0.7) R Z All ’ have pulls of ~1, all the B have pulls of ~1.2.. There seems to be no significant dependence except from ct!!!  Pt

  30. Backup

  31. Outline • Current status • What was used for the mixing results • What is the current understanding of Ebe • Plans for improvements • How can we improve?

  32. Current status EbE: itearative track selection/pruning algorithm to provide an unbiased estimate of the PV position on an Event-by-Event basis • Hadronic analyses used a flat ~25um beamline! • Possible improvements: • Move to “hourglass” • Move to EbE • EbE + Hourglass • One of the ½ leptonic analyses used this with fixed hourglass parameters Hourglass

  33. What do we know about EbE? • Unbiased estimator of PVTX Reasonable (~5%) control of systematics

  34. Cross checks using I.P.(B) • Lxy involves three ingredients: • EbE • Secondary vertex • Beamline (in beamline constrained fits) Something funny when beamline is used! Scale factors work! Z dep. Beamline improves pulls! B Lxy d0

  35. Time dependence of Hourglass parameters Implementing DB access of time-dependent parameters

  36. What do we gain? Euphemism • 15-20% In vertex resolution! • Better control of systematics (hard to evaluate) • Correct EbE resolution (it is not clear that it is correct now) • Red arrow is the effect of 1. Only • Point 2. Affects mostly the green area (tiny ?) • Point 3. Has an effect qualitatively similar to 1., but hard to evaluate

  37. Hadronic analysis systematics ct scale factor 0.000 0.024 0.061 0.090 0.144

  38. Hourglass parameters from DBProfiles

  39. B Relative PV/BV contribution to d0 and Lxy pulls w w Lxy d0 • PV and BV are linear combinations of the same covariances (PV, SV), with different coefficients • Lxy sensitive to the major axis of SV • Relative weight of PV and SV covariances different for Lxy and d0 • Look at: Note:the two Lxy (or d0) pieces do not linearly add to 1!

  40. B Relative PV/BV contribution to IP and Lxy pulls Lxy d0 • Not Beam Constrained • Beam constrained • Beam constrained with run-dep. hourglass

  41. K K* Lxy(2) Pull K+ Lxy(2) Pull Two gauss. One Gaussian

  42. Charm… ’ “3-1” Lxy(2) Pull D+ Lxy(2) Pull Two gauss. One Gaussian

  43. ’ can be used in two different ways to probe SV ’ “2-2” Lxy(2) Pull “3-1” Two gauss. One Gaussian “2-2”

  44. Bottomline: • SV and PV enter very differently in Lxy and d0 • Relative contribution depends strongly on PV and SV scales • Beam constraint squeezes the PV resolution significantly. Becomes second order on Lxy! • We are in a regime where the SV scale factor is critical! … now let’s get more quantitative!

More Related