Fast TracKing at ATLAS Why and How Jamie Saxon University of Chicago
What is the FTK? ‣ Global tracking early in the HLT (100 µ s). (p T , η , φ , d ₀ , z ₀ , hits, and χ² ) ‣ Dramatic step for the ID in the trigger. ‣ Currently in production and installation. ‣ Operating next year through 2022.
Why do this? A. Lower thresholds for b, τ triggers. ‣ Tracking currently limited to cones around L1-objects. ‣ 3rd gen very interesting for Run II. B. Confront pileup through vertexing. ‣ Good for pileup corrections, multijet triggers, and Ɇᵀ corrections/track Ɇᵀ . C. Exotic: displaced vertices, rare flavor, unbiased sample (large µ ), ‘mini-events.’
How? (Algorithmically) φ 2 φ 1
How? (Algorithmically) φ 2 φ 1 φ 2 Not a Track Maybe!? Track Not a Track φ 1
How? (Algorithmically) φ 2 φ 1 1. Find patterns of plausible candidates. φ 2 Not a Track Maybe!? Track Not a Track φ 1
How? (Algorithmically) φ 2 φ 1 1. Find patterns of plausible candidates. 2. Evaluate χ² , linearized in small detector regions. φ 2 Maybe!? Track φ 1
How? (Algorithmically) φ 2 φ 1 1. Find patterns of plausible candidates. 2. Evaluate χ² , linearized in small detector regions. φ 2 3. ‘Guess’ missing hits. Maybe!? Track φ 1
How? (Algorithmically) φ 2 φ 1 1. Find patterns of plausible candidates. 2. Evaluate χ² , linearized in small detector regions. φ 2 3. ‘Guess’ missing hits. Maybe!? 4. Linearized calculation of Track helix parameters. φ 1
At ATLAS… ‣ Split pixel and SCT hits from the detector into 64 parallel pipelines . Silicon Hits HLT Photos: atlasftk.uchicago.edu/photos/
Data Formatter ‣ Cluster hits, share them among regions. ‣ Finite beamspot, curvature at low-p T . Silicon Hits Data Formatter
Auxiliary Card ‣ Coarse resolution hits to pattern matching. Silicon Hits Auxiliary Data Card Formatter
Associative Memory Board ‣ Find coarse-resolution track candidates. ‣ Massive fanout to 64 custom chips. Silicon Hits Auxiliary Data AM Chip 06 Card Formatter ‣ Loads hits in memory. ‣ All candidates matched simultaneously (magic). Assoc. ‣ Variable-width patterns. Memory ‣ Optimized for low-power. Board
Auxiliary Card ‣ Linearized χ² for patterns matched in the AMB, using 8 detector layers. Silicon Hits Auxiliary Data Card Formatter Assoc. Memory Board
Second Stage Board ‣ Extrapolate aux. fit to 12 layers. ‣ Calculate track parameters, new χ² . 4 silicon layers Silicon Hits Second Auxiliary Data Stage Card Formatter Board Assoc. Memory Board
FTK-Level 2 Interface Card ‣ Restore ‘global’ hit coordinates. ‣ Package complete events for HLT. FTK- Silicon Hits Second Auxiliary Level 2 Data HLT Stage Interface Card Formatter Board Card Assoc. Memory Board
The System ‣ 322 9U boards, 640 mezzanine cards, 7 racks… 250B fits/second. Silicon Hits HLT
Tracking Performance ‣ E ffi ciencies WRT o ffl ine > 90%. ‣ Excellent resolution for η , φ , d ₀ , z ₀ . ‣ Good p T resolution; ‘cheap’ to refit. 1 Resolution Efficiency WRT Offline 0.95 at 10 GeV 0.9 p T 0.5 GeV 0.85 0.8 η 0.0015 0.75 0.7 φ 0.001 0.65 muon 0.6 d ₀ 0.06 mm pion 0.55 ATLAS Simulation, no IBL 0.5 z ₀ 0 10 20 30 40 50 60 70 80 90 0.25 mm p [GeV] FTK TDR: cds.cern.ch/record/1552953 T
Physics Performance ‣ Great discrimination on physics quantities, e.g., σ (d ₀ )! Normalized Entries ATLAS Simulation, no IBL Barrel (| | < 1.1) η < > = 69 µ -1 10 Offline Light-Flavor Offline b-Jet FTK Light-Flavor -2 10 FTK b-Jet -3 10 -4 10 -5 10 -20 -10 0 10 20 30 40 d0 Significance FTK TDR: cds.cern.ch/record/1552953
Trigger Impact: τ ’s and b’s ‣ E ffi ciencies v. rates for b and τ triggers, WITH and or WITHOUT FTK: significant gains. ‣ What can you do with global tracking? b -Jet Triggers Tau Triggers FTK TDR: cds.cern.ch/record/1552953
Take-Away ‣ Starting next year, FTK will provide global tracking for every level 1 trigger at ATLAS. ‣ Good resolution of track parameters. ‣ > 90% e ffi cient WRT o ffl ine. ‣ Threshold-less b and τ -tagging, PV information, and ‘exotic’ applications. ‣ Exciting time: intellectual challenges, hands-on work, and physics potential now .
Recommend
More recommend
