The Run-2 ATLAS. ATLAS Trigger System: Design, Performance and Plans

Transcription

1 The Run-2 ATLAS Trigger System: Design, Performance and Plans 14th Topical Seminar on Innovative Particle and Radiation Detectors October 3rd October 6st 2016, Siena Martin zur Nedden Humboldt-Universität zu Berlin for the ATLAS Collaboration Page 1

2 Overview ATLAS Trigger Overiew Level-1 and Higher Level Trigger Performance and Improvement Upgrade for Run-2 Trigger Menus Trigger Objects electron, muon, tauon JETs, missing transverse energy combinations Further Developments for Run-2 L1 topological Trigger L1Topo FastTracK-Trigger FTK Page 2

3 ATLAS Trigger Overview Trigger Functionality Trigger Menu Physics Yields Page 3

4 Introduction: LHC in 2016 LHC: very dense environment Bunch crossing rate of 40 MHz L > 1.2 * 1034 cm-2s-1 ~ 23.2 interactions per bunch crossing Sophisticated Trigger system High selectivity (purity and efficiency) needed Crucial element of ATLAS system Reliable early data reduction crucial for physics program Balance between low thresholds, hight rates, limited bandwidth and online computing resources Complex selection strategies Challenging environment for Trigger Operation Very successful in Run-1 Many improvements during shutdown Very efficient and fast re-comissioning for Run-2 Stable and reliable running in Run-2 Page 4

5 Schematic Overview L1 Input: 40 MHz Output: ~ 100 khz HLT Input: ~ 100 khz Output: ~ 1 khz physics (full event building) ~ 2 khz calibration, monitoring and Trigger level analysis (partial event building) L1Topo in commissioning phase (running) FTK in test phase O(40k) Page 5

6 Schematic Overview: L1 L1 Input: 40 MHz Output: ~ 100 khz 40 MHz 100 khz, 2.5 μs latency 512 decission items Fast, custom made electronics Find RoIs based on muon and calorimeter data RoI: Region of Interest in η/φ RoIs handed over to HLT Upgrades: L1Calo, L1Muon, L1CTP New systems: L1Topo, FTK HLT Input: ~ 100 khz Output: ~ 1 khz physics (full event building) ~ 2 khz calibration, monitoring and Trigger level analysis (partial event building) Page 6

7 Schematic Overview: HLT Large Farm with O(40 k) processors ~ 2500 independent trigger decision chains Event building at a flexible time within the chain Average latency: ~ 200 ms Fast and custom HLT algorithms offline-like algorithms within the L1 RoIs Full event reconstruction also possible Upgrades for Run-2: Upgrade of readout and data storage systems Unification of Level-2 and Event-Filter trigger levels to one single Higher Level Trigger optimized use of farm resources L1 Input: 40 MHz Output: ~ 100 khz HLT Input: ~ 100 khz Output: ~ 1 khz physics (full event building) ~ 2 khz calibration, monitoring and Trigger level analysis (partial event building) O(40k) Page 7

8 Trigger Strategy L1 RoIs seed trigger chains at HLT HLT writes to data streams Trigger Menu: configures the full decission process Connection of L1 items HLT chains streams Appropriate prescales at L1 and HLT on top Main physics stream: Primary triggers Support, alternative and backup triggers Calibration and monitoring triggers Menu composition Chain: sequence of selection algorithms with increasing complexity and detector information L1 item: RoI with a certain threshold from Calo or Muon High rate, but only partial event building High acceptance for physics (BSM / Higgs / SM / top physics / ) Calibrattion and Monitoring Respect the limits of the TDAQ system Largest bandwidth consumption: Main physics stream Page 8

9 Trigger Groups Trigger Menu composition optimized for several luminosity points Additional triggers and decreased prescales for lower luminosity points to make maximal use of bandwidth Change the trigger composition at predefined luminosities L1: MU / EM / JET / MET / TAU HLT: mu, electron, photon, jet, b-jet, MET, tau, B-Physics Combined and multiple objects at both levels Balance between thresholds, multiplicities and prescales: highest possible physics output Ensure the system stability (avoidance of dead-time) Page 9

10 Trigger Menu Very rich physics program of ATLAS: Large variety of trigger chains Primary triggers unprescaled Support triggers prescaled Organized according to Trigger Objects (TOB): e / μ / γ / τ / jets / b-jets / Etmiss Various thresholds, multiplicities and combinations Optimized for specific luminosities Dedicated menus for various running conditions Maximum expected luminosity pp and Heavy Ion running Special runs (e.g. VdM scan) Page 10

11 ATLAS Trigger Performance Electron / photon trigger Tau trigger Top, Higgs, SM Missing transverse energy trigger Higgs, BSM Jet (b-jet) trigger Higgs, top (electron), SM, SUSY BSM, SUSY Muon trigger Top, SM, B-physics, SUSY Page 11

12 Electron, Photon and Tau Trigger e: likelihood based identification γ: cut based identification τ: BDT based indentification (hadronic trigger) Efficiency turn-on as a function of reconstructed E T/pT Steep turn-on curves: No waste of bandwidth, plateau at ~100 % Comparison to MC: good agreement Tag-and-probe method with Z ee / ττ Page 12

13 Jet and MET Trigger Jet-trigger improvements for Run-2 Jet Trigger uses now entire calorimeter data with anit-kt algorighm Possible thanks to algorithm improvements: Pile-up mitigation is the main challenge for ETmiss triggers ~ 7 x faster data unpacking, ~ 2 x faster clustering mht algorithm based on pt sum of HLT jets currently default Efficiency shown as a function of offline jet-p T / offline ETmiss Sharp efficiency turn-on curves in large jet-pt / ETmiss ranges Page 13

14 Muon Trigger Improvements for Run-2 New coincidence logic at L1 suppresses fake triggers in forward region from muon trigger Coverage improved with installation of new chambers 100 % efficiency of HLT wrt. L1, sharp turn-on Page 14

15 ATLAS Trigger new Developments L1 Topological Trigger L1Topo Fast TracKer FTK Page 15

16 New developments: L1Topo Limitations at higher / increasing luminosity: Higher trigger rate due to higher pile-up Need to down-scale rates and to increasing thresholds reduced physics acceptance L1Topo crucial for making use of the luminosity upgrade Topological calculation on L1 with Trigger Objects from L1Calo and L1Muon Up to 128 configurable algorithms (FPGA) Combining calorimeter and muon L1 trigger Selecting on complex physics quantities Angular distances: η, φ, Δη, Δφ, ΔR Missing, transverse and invariant mass calculations Compound triggers e/γ, jets, μ, τ, ETmiss, HT = Σ(pT) Physics oriented system Trigger for specific physics channels Page 16

17 Commissioning of L1Topo Based on a calorimeter trigger sample using HT triggers Efficiency turn-on curve as a function of the reconstructed HT (scalar sum of jet-pt) Rates as a function of the luminosity for two HT triggers Online measured rates compared to prediction from trigger simulations Good progress, L1Topo can soon be used in the selective mode Page 17

18 New Developments: FTK Fast TracKer: Hardware-based track finder Performs a global track reconstruction (pt > 1 GeV) from silicon trackers Provides full-event track information to the HLT at full Level-1 rate (100 khz) using associate memory for pattern matching and FPGAs for track fitting Pile-up robustness and secondary vertex finding for heavy flavour physics Installation and tests ongoing Page 18

19 Summary and Outlook Run-2 requirements from LHC are a real challenge for the trigger system ATLAS trigger system in very good shape Major improvements and upgrades for Run-2 working fine Well-designed and working trigger menu Meeting the physics performance requirements Extensive studies of the Trigger performance Further developments ongoing L1 topological trigger in commissioning phase Fast Track Trigger in testing phase Page 19

20 Backup HLT physics group performance plot Tau-Trigger performance MET-Trigger performance B-Physics Trigger Page 20

21 Trigger Menu: HLT Gropus L1: MU / EM / JET / MET / TAU HLT: mu, electron, photon, jet, b-jet, MET, tau, B-Physics Page 21

22 Tau Trigger Hadronic trigger: looking at hadronic tau decays τ: BDT based indentification Efficiency turn-on curve measured in data as function of the offline transverse momentum p T measured in a tag and probe analysis with Z ττ μτhad Comparison to simulations and to L1 / HLT Page 22

23 MET Trigger MET: Trigger on Missing Transverse Energy Pile-up mitigation is the main challenge for ETmiss triggers mht algorithm based on pt sum of HLT jets currently default Efficiency as a function of the offline ETmiss: Selected from single e/μ triggers Seeded by L1 ETmiss > 50 GeV Page 23

24 B-Physics Trigger Di-μ trigger: 4 GeV+6 GeV candidates at L1, confirmed at HLT Single-μ trigger: 15 GeV at L1, 20 GeV threshold at HLT Without further requirements very high rate with increasing luminosity, large prescales would be needed Usage of L1Topo vital to reduce the di-muon trigger rates keeping the physics program: Clear reduction of the rate compared to the 2MU only trigger visible Signal sample: B0s μμ Page 24