FTK for the τ Trigger Phillip Urquijo Soshi Tsuno Atlas Trigger Workshop Athens December 3, 2012
Introduction • The FTK will improve τ triggering. underlying • Primary vertex information allows tau event pile-up • τ -jet vertex association • More accurate perigee for Δ z 0 and pileup robustness • Unified selection at L2 • tau triggering sequence is in 3 steps, count calculate limiting effectiveness of MV selection # tracks R EM , R track • τ selection efficiency in cone in cone • Improved track cluster shape resolution. 0.2 0.4 ∆ R • PV reweighting of calo. based selection. FTK in τ triggers 2 Phillip URQUIJO
Fraction of track p “Jet V from the p PV p T (track) � vertex. all p T (track) � PV a m l i n e b e !"#$%"&'()"*)+ !#,)-.!'()"*)+ !#,)-.!'()"*)+ Ryan Reece (Penn) 1. Vertex association 2. IP resolution
2012 Pileup dependence & Tracking • Most pileup dependence in 2011 was due to Energy- weighted calorimeter radius variables using the EM cal. • Remedied in 2012 in 3 ways: • Smaller cone sizes. • Removed most sensitive calo. criteria: variables in τ recon. now based on track info. • Track shapes based on association/distance to highest p T track ( Δ z 0 ) - pseudo PV reference . FTK in τ triggers 4 Phillip URQUIJO
Remaining Pile-up Dependence • Δ z 0 is effective but imperfect. In 2012 “Jet Vertex Fraction” raction of track p T = from the primary JVF(jet , vertex) some loss w.r.t. offline τ , partly due to ( ) vertex. � p T (track) tracking/PV: � � tracks matched • RoI limitations on track finding to jet and vertex ( ) • PV location (with TJVA ), for IP & flight � � p T (track) � tracks matched • nVtx reweighting of calo. thresholds to jet • Δ z 0 criterion reduces separation 10 power for τ ID. FTK in τ triggers 5 Phillip URQUIJO
Tau-jet vertex association: Offline • TJVA is used offline: τ τ Number of selected Number of selected 0.8 • Choose vertex with highest 0.6 0.7 =0 µ µ =0 0.5 JVF for that τ candidate. =20 µ 0.6 =20 µ =20 with TJVA µ =20 with TJVA µ 0.4 0.5 • Recovers N track bin ATLAS Internal ATLAS Internal Simulation Simulation 0.4 0.3 0.3 migration - important 1P MP 0.2 0.2 since cuts are N track 0.1 0.1 dependent. 0 0 0 1 2 3 4 0 1 2 3 4 Number of tracks Number of tracks • Largest effects at low (a) True 1-prong taus (b) True 3-prong taus track p T at high pileup . 1 1 S S E E 0.95 0.95 • Need good IP resolution 0.9 0.9 0.85 0.85 for low p T τ tracks. =0 µ 0.8 0.8 =0 µ =20 µ 0.75 0.75 =20 µ =20 with TJVA µ =20 with TJVA µ Parameters f JVF 0.7 0.7 ATLAS Internal ATLAS Internal Simulation 0.65 0.65 Simulation ∆ Z ( trk , v tx ) 2mm 1P MP 0.6 0.6 ∆ Z ( trk , v tx ) / σ ( Z trk ) 1000 0.55 0.55 0.5 0.5 d 0 2mm 0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40 Tau charged pion p [GeV] Tau charged pion p [GeV] T T d 0 / σ ( d 0 ) 1000 (a) True 1-prong taus (b) True 3-prong taus FTK in τ triggers 6 Phillip URQUIJO
Improved perigee for Δ z 0 : Trigger • d 0 &z 0 depend on choice Number of tracks s 12000 0.35 � (0, 0, 0) Number of TJVA TJVA of PV: 0.3 ex00 10000 ex00 exBS exBS • 4 techniques examined 0.25 8000 0.2 for trigger (K.G. Tan): 6000 0.15 1.Default 4000 0.1 2.Recalculate perigee at 2000 0.05 (0,0) 0 0 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 z (wrt true PV) [mm] � IPz (wrt lead track) [mm] � pos 3.beam spot BS 0 4.TJVA 1 1 Track selection efficiency Track selection purity • TJVA provides well 0.9 0.9 0.8 0.8 resolved dz 0 . 0.7 0.7 0.6 0.6 • BS will provide similar 0.5 0.5 |TJVA IPz |<1.5mm 0 |Default IPz |<2mm 0.4 � � 0 0.4 performance of |ex00 IPz |<1mm No IPz cut � � 0 0 0.3 |ex00 IPz |<1.5mm |TJVA IPz0|<1.5mm 0.3 � 0 |Default z |<2mm � Efficiency/Purity if 0 � 0.2 |ex00 � z |<1mm 0.2 0 |ex00 IPz0|<1.5mm 0.1 0.1 accurate (offline BS used 0 0 0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40 for this study). µ µ FTK in τ triggers 7 Phillip URQUIJO
Improved perigee for Δ z 0 : Trigger • d 0 &z 0 depend on choice Number of tracks s 12000 0.35 � (0, 0, 0) Number of � TJVA TJVA of PV: 0.3 ex00 10000 ex00 � exBS exBS • 4 techniques examined 1 0.25 8000 � 0.2 for trigger (K.G. Tan): 0.9 6000 � 0.15 1.Default 4000 0.1 0.8 2.Recalculate perigee at 2000 0.05 0.7 0 (0,0) 0 0 -3 -2 -1 0 1 2 3 0 5 10 15 20 25 30 35 40 -3 -2 -1 0 1 2 3 z (wrt true PV) [mm] � IPz (wrt lead track) [mm] � pos 3.beam spot BS 0 µ 4.TJVA 1 � 1 Track selection efficiency Track selection purity • TJVA provides well 0.9 0.9 � 0.8 0.8 resolved dz 0 . 0.7 0.7 � 0.6 0.6 • BS will provide similar � 0.5 0.5 |TJVA IPz |<1.5mm 0 |Default IPz |<2mm 0.4 � � 0 0.4 performance of |ex00 IPz |<1mm No IPz cut � � 0 0 0.3 |ex00 IPz |<1.5mm |TJVA IPz0|<1.5mm 0.3 � 0 |Default z |<2mm � Efficiency/Purity if 0 � 0.2 |ex00 � z |<1mm 0.2 0 |ex00 IPz0|<1.5mm 0.1 0.1 accurate (offline BS used 0 0 0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40 for this study). µ µ FTK in τ triggers 7 Phillip URQUIJO
EF/Offline BDT comparison 1. Track variable resolution 2. PV weighting
Selection at EF and Offline • 2012 selection largely track based , calorimeter quantities are sensitive to pileup. (dZ 0 requirement). FTK could help with 1. nVtx dependent calo threshold reweighting, and 2. track cluster shape & IP resolution 1P 1P MP MP Variables Variables ariables ariables code name EF Off. EF Off. R track ⚈ ⚈ ⚈ ⚈ trkAvgDist 〉 0.22 EM ATLAS Preliminary m tracks ⚈ ⚈ R 0.2 massTrkSys 〈 jets 0.18 S Tflight ⚈ ⚈ trFlightPathSig 0.16 Track Track S lead track 0.14 ⚈ ⚈ p u ipSigLeadTrk - e l i p 0.12 N widetrack ⚈ ⚈ true τ had nWideTrk 0.1 0.08 Δ R max ⚈ ⚈ dRMax 0.06 Calo. f core 0.04 ⚈ ⚈ * ⚈ ⚈ * centFrac 2 4 6 8 10 12 14 N(vertex) Calo. N vtx +Track f track ⚈ ⚈ * ⚈ ⚈ * etOverPtLeadTrk ⚈ *: Pileup corrected variables FTK in τ triggers 9 Phillip URQUIJO
Pile up corrected variables (Offline) Δ R< 0.4 r E T ,j ] ∆ R j < conesize Δ R< 0.2 j ∈ all X f core linear fit: slope: 296 slope: 74 2011 Ztautau MC nVtx N vtx y f core on f’ core P ∆ R i < 0 . 1 E T ,i i ∈ all f core = P ∆ R j < 0 . 2 E T ,j f core j ∈ all linear fit: slope: -0.006 N Vtx pileup corrected slope: -0.000 2011 Ztautau MC f 0 core = f core + 0 . 006 · N vtx nVtx N vtx FTK in τ triggers 10 Phillip URQUIJO
Offline - EF comparison of Track shapes • Room for improvement in EF resolution of track observables, if PV, and RoI limitations can be avoided (A. Tanasijczuk) • Track variables with poor EF resolution R track N widetrack • µ=30 • IP dependent variables. S Tflight S lead track FTK in τ triggers 11 Phillip URQUIJO
Sensitive variables (Not currently used) • Many variables not used in selection in 2012 Normalised events 0.12 ATLAS Preliminary offline and online, due to pile up sensitivity. Signal ! MC offline 0.1 • Although offline could deal with it, offline Signal ! MC EF Dijet Data (2011) offline 0.08 Dijet Data (2011) EF BDT simplified for harmony with trigger. 1-prong 0.06 Rejection power not optimal but stable 0.04 efficiency. 0.02 • With PVs (multiplicity&positions) can apply 0 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 R EM nVtx dependent cuts to e.g. EM calo. radius. • e.g. Variables not (directly) to remedy pileup dependence. used ... • MVA training can be performed to be flat in • Calo-radius (EM&Had) pT & nVtx • Topocluster-invariant mass • Then we can consider more calo. variables. • Ratio of Isolation to Core • Preselection reduces rejection power: Smaller Scalar P T Sum • Electron veto (TRT HT) ∆ R for R EM 0.4 → 0.2, ∆ z 0 at 2mm. So it could also be revisited. • Stripwidth FTK in τ triggers 12 Phillip URQUIJO
Unified L2 selection Use FTK to remove calo preselection stage from L2
Selection at L2 & Multivariate feasibility • Unlike EF & Offline, selection is 3 steps ε (FEX+HYPO) 3 . L2 MV algorithms severely limited: rejection factors kept low to minimise efficiency loss (L2 Calo. energy resolution is poor, and not ideal for a first step of selection ) • Limiting factor for a 1 step MV is tracking timing. tableChnL2 L2_tau29_medium_L2StarB calls_cpu_time ATLAS data 2012 L2StarB Step Variable 3 2stTest 10 EtRawMin 1. L2Calo 1. L2Calo Algorithm calls CoreFractionMin 2 10 NtrkMax 2. L2Track 2. L2Track 10 SumPtRatioMax EtOverPtLeadTrkMax 3. L2Tau 3. L2Tau 1 TrkAvgDistMax 0 10 20 30 40 50 60 70 80 90 100 Algorithm CPU time (ms) • 2012: Largely track based, calorimeter quantities are sensitive to pileup. (dZ 0 requirement). • Rejection ~ 10, c.f . >100 at EF. FTK in τ triggers 14 Phillip URQUIJO
Recommend
More recommend
