Vertexing Degradation Takanori Hara(Osaka U.) 2005/Apr/21 @Hawaii
Near Future 2004 2008 Lumi. 13/nb/s 60/nb/s HER 1.2 A 2 A LER 1.8 A 3 A SVD nominal x1.7~2.6 CDC nominal x1.7~2.5 ~x3 BG study is needed
What will happen? Red : x3 BG 1st 2nd 3rd 4th lyr. Blue : nominal ~25% SVD x3 BG Occupancy will increase upto ~25% in 1st layer! ~8% Vertexing is OK?! Strip #
Performance Prediction Procedure B J/ Ψ Ks, B generic MC event (Detector Simulation) + Random-trigger real Background event (as of 2004) Nominal BG x3 BG Tracking/Vertexing Performance Check Some Assumptions PID is perfect, Lepton/Kaon tagging only, Current Hardware/Software
Vertex Resolution ( σ = single gaussian) Nomial x3 BG σ ~36 µ m σ ~44 µ m MC Rec ∆ V cp ∆ V cp (=V cp - V cp ) σ ~67 µ m σ ~79 µ m MC Rec ∆ V tag (=V tag - V tag ) ∆ V tag σ ~86 µ m σ ~104 µ m n o i t MC Rec a ∆ V dif d ∆ V dif (=V dif - V dif ) a r g e d % 0 2 ~
Reconstruction Efficiency Nomial CP side x3 BG Track finding in CDC % ~3% 100 Track Quality(SVD+CDC) (cp vtx/L0 trig) 80 L0 trig Nominal : ~67% ~10% 60 4 part. x3 BG : ~58% cp vtx TAG side 40 Track finding + Quality mass cut tag vtx (J/ Ψ ,Ks,B) (tag vtx/cp vtx) 20 ~3% tag Nominal : ~49% 0 x3 BG : ~46%
Most Responsible Part? B J/ Ψ Ks, B generic : 10,000 events ( σ = single gaussian) 1st : Nominal 1,2 : Nominal 1st : x3BG 2,3,4 : x3BG 3,4 : x3BG 2,3,4 : Nominal ~36 µ m ~36 µ m ~44 µ m ∆ V cp 3321 events 3382 events 3381 events ~66 µ m ~66 µ m ~72 µ m ∆ V tag 1531 events 1558 events 1568 events ! ~89 µ m ~85 µ m ~99 µ m ∆ V dif r e y a resolution is not so different l t s 1 from nominal BG case
SVD Cluster Classification (S, S) True cluster S (B, B) Background(BG) cluster Signal B Fake cluster (B,S), (S,B) Background S+B S Signal (S+B,S) Distorted cluster Background B BG/Fake/Distorted clusters smear hit position affect tracking/vertexing B S
CP side(J/ Ψ K s ) + - CP vertex = Vertex ( ) J/ Ψ µ µ x3 BG, 1st layer µ momentum Red : True cls energy diff. of 1st lyr Blue : Distorted Green : Background Yellow : Fake |E z |-|E r φ | (elec) P(GeV/c) Distorted cluster is the main cause
Distorted Cluster 2149 1st lyr. Distorted cls. ~80% 425 1724 2nd lyr. hit (True) NO hit ~60 ~360 3rd lyr. NO hit hit (True) hit (True) 4th lyr. hit (True) hit (True) hit (True) Tracking seems to be fine upto 2nd layer Clustering sould be improved !
TAG side x3 BG, 1st layer µ momentum Red : True cls energy diff. of 1st lyr Blue : Distorted Green : Background Yellow : Fake |E z |-|E r φ | (elec) P(GeV/c) Fraction of BG/Fake cluster increases
Background Cluster 2985 1st lyr. Background cluster 1143 1842 2nd lyr. hit NO hit (BG ~700) (True ~200,Fake~250) ~800 ~1000 3rd lyr. NO hit hit (True ~300) NO hit (BG+Fake ~500) ~900 4th lyr. hit NO hit NO hit (True ~300) (BG+Fake ~550) 2 of 3 remaining layers have no hit. Track seed itself in CDC is not good !
Fake Cluster 2354 1st lyr. Fake cluster 1038 1316 2nd lyr. hit NO hit (True ~650,Fake~300) ~700 ~600 ~300 3rd lyr. NO hit hit (True ~300) NO hit hit (True ~650) (Fake ~200) ~700 4th lyr. hit NO hit:hit hit NO hit :hit (True ~100) (Fake ~550) (2:3) (True ~600) (1:1) Good track Poor track seed ? + poor cluster in 1st lyr ? (worsens the track quality)
+ B D* D* Case + e + - g B D* D* (D* D(K π ) π ), B generic - n a r x3 BG, 1st layer m u t n mom. of CP-side part. mom. of TAG-side part. e m o m e m a S Red : True Blue : Distorted Green : Background Yellow : Fake Not appropriate P(GeV/c) P(GeV/c) CP side High mom. TAG side Low mom.
Points so far Causes of the vtx-resolution degradation High Momentum (Distorted cluster) Tracking (upto 2nd lyr) is fine Distorted cluster in 1st lyr worsens trk-quality Low Momentum (Background cluster) Poor seed track in CDC Low Momentum (Fake cluster) Poor seed track in CDC (~a half) Fine tracking (upto 2nd lyr) +Bad cluster in 1st lyr.
How to Recover ... Current Attempt Energy balance cut ||E z |-|E r φ ||> 15,000 el Cut cls energy diff. (CP side) Not used for tracking in SVD (for 1st, 2nd layer) Cut Cut Quality of track w/o Bad clst > w/ Bad clst (esp. 1st lyr) (esp. 1st lyr) Better vtx resol. |E z |-|E r φ | (elec)
Recovered Vtx Resolution Better vtx resol. B J/ Ψ Ks, B generic : 10,000 events ( σ = single gaussian) x3 BG Nominal BG x3 BG w/ ene. balance ~36 µ m ~44 µ m ~42 µ m ∆ V cp 4197 events 3115 events 3120 events ~75 µ m ~67 µ m ~79 µ m ∆ V tag 1423 events 2038 events 1436 events ~86 µ m ∆ V dif ~104 µ m ~96 µ m
Other Attempts... For High-momentum tracks To avoid using Distorted clsuter in the 1st layer Is it possible to optimize clustering ? or cluster selection ? (by Heffernan) 1st lyr. 1BG, N-side cluster width, 2-strip clustering thresold, incident angle of particle, 3-strip 1-strip layer correlation of charge ..... 4-strip 5-strip WIP incident angle(deg)
Other Attempts... For Low-momentum tracks To get more reliable low-momentum tracks SVD self-track finding is difficult ... (SVD 4 lyrs + sCDC) Eff. (by Trabelsi@U of Hawaii) } 200MeV/c 50MeV/c SVD+sCDC track finding ? Still working on
Possible Hardware Updates Optimizing shaping time of readout chip in SVD (Background ~30%@max ) Or replace the readout chip on the 1st layer VA1TA APV25 (shaping time 800nsec ~50nsec) Pixel-type detector can reduce Distorted/Fake cluster Replace the readout electronics in CDC Deadtime can be reduced (2.2 µ 0.8 µ sec) ..... Key point is to reduce room for BG to sneak in
Summary Vertex Resolution deteriorates by ~20% under x3 BG Causes of deterioration for High mom. track Track finding/quality (upto 2nd lyr) seems to be OK Distorted cluster in 1st lyr makes track quality worse for Low mom. track Track seed itself in CDC is poor SVD true cluster can not be used correctly Way of Recovery Better vertex resolution Energy balance cut But many things to do (low-pt tracking, better clustering, hardware...)
Reconstruction Efficiency ( π + π -) CP side Track finding in CDC % ~1.4% 100 Track Quality(SVD+CDC) (cp vtx/L0 trig) 80 Nominal : ~88% ~6% 60 x3 BG : ~82% TAG side 40 Track finding + Quality (tag vtx/cp vtx) 20 ~1% Nominal : ~48% 0 x3 BG : ~47%
CP side( π + π -) + - CP vertex = Vertex ( ) B π π x3 BG, 1st layer π momentum cls energy diff. of 1st lyr Red : True Blue : Distorted Green : Background Yellow : Fake |E z |-|E r φ | (elec) P(GeV/c) Situation is the same. Distorted cluster !
Recovered Vtx Resolution + - B π π , B generic : 10,000 events ( σ = single gaussian) x3 BG Nominal BG x3 BG w/ ene. balance ~26 µ m ~32 µ m ~30 µ m ∆ V cp 7023 events 5775 events 5867 events ~71 µ m ~59 µ m ~72 µ m ∆ V tag 2754 events 3356 events 2725 events ~71 µ m ~97 µ m ~88 µ m ∆ V dif Better Efficiency ?! Better Vtx Resol.
Recommend
More recommend
