RAL Summer SchoolSeptember, 2004
Efstathios (Stathis) [email protected] http://www.hep.ucl.ac.uk/~sstef
High Level Trigger Studies for the
ATLAS Detector
Efstathios (Stathis) Stefanidis
University College London
RAL Summer SchoolSeptember, 2004
Efstathios (Stathis) [email protected] http://www.hep.ucl.ac.uk/~sstef
OUTLINE The ATLAS Detector : Overview.
The Trigger System.
i. Overview.
ii. High Level Trigger.
Performance Studies.
i. IDScan.
ii. e/γ vertical slice.
iii. RoI size.
Future Plans.
RAL Summer SchoolSeptember, 2004
Efstathios (Stathis) [email protected] http://www.hep.ucl.ac.uk/~sstef
The ATLAS Detector: Overview.1. pp colliding beams
14 TeV c.m. energy.
2. Design Luminosity: L=1034 cm-2s-1.
3. Three Parts:
a) Inner Detector
b) Calorimeter (EM – HAD)
c) Muon Spectrometer
4. Magnet System:
a) Solenoid: 2 T
b) Toroid: 0.4 T
RAL Summer SchoolSeptember, 2004
Efstathios (Stathis) [email protected] http://www.hep.ucl.ac.uk/~sstef
The Trigger System: Overview.1. LVL1:
• 40 MHz 75 kHz • < 2.5 μs • Hardware trigger• Reduced granularity information
2. LVL2: • 75 kHz 2 kHz • 10 ms • Full granularity information
from both ID and Calorimeter• RoI mechanism
3. Event Filter (EF): • 2 kHz 100 Hz 2s • Sophisticated algorithms• Alignment data available
High Level Trigger (HLT)
RAL Summer SchoolSeptember, 2004
Efstathios (Stathis) [email protected] http://www.hep.ucl.ac.uk/~sstef
The Trigger System: HLT.
• RoI mechanism
(i) Defines the area where HLT
will start from.
(ii) Seeded by LVL1
(iii) Only the data needed are
transferred
MINIMISE THE PROCESSING TIME AND NETWORK TRAFFIC
Objects Selection Signatures
Physics Coverage
Electron e25i, 2e25i Higgs (SM, MSSM), new gauge bosons etc
Jet j400, 3j150 SUSY, compositeness, resonances etc.
1. Event Selection Strategy:
• Signature validation
Checks for signatures coming from interesting physics events
REJECT UNINTERESTING EVENTS VERY EARLY
RAL Summer SchoolSeptember, 2004
Efstathios (Stathis) [email protected] http://www.hep.ucl.ac.uk/~sstef
The Trigger System: HLT.2. Event Selection
Software:
• ATHENA
(i) Written using GAUDI architecture
(ii) Provides common services (Transient Data Store, Histograms, Auditing etc)
(iii) Well-defined interface not only to developers but also to the end-users
IMPROVE COHERENCY OF THE DIFFERENT SOFTWARE DOMAINS
• IDScan
(i) LVL2 track reconstruction algorithm
(ii) SPACE POINTS as input – Track parameters as output
(iii) Runs several times per event and once per RoI
RAL Summer SchoolSeptember, 2004
Efstathios (Stathis) [email protected] http://www.hep.ucl.ac.uk/~sstef
Performance Studies: IDScan.Pattern Recognition
Algorithms Low Luminosity
High Luminosity
IDScan (89.0 ± 0.4) %
(82.9 ± 0.5) %
xKalman (88.9 ± 0.4) %
(84.0 ± 0.5) %
iPatRec (87.2 ± 0.4) %
(82.2 ± 0.5) %
|pTgen – pT
rec|<15 GeV
|ηgen – ηrec|<0.01
|φgen – φrec|<0.01 rads
RAL Summer SchoolSeptember, 2004
Efstathios (Stathis) [email protected] http://www.hep.ucl.ac.uk/~sstef
Performance Studies: e/γ slice.•Determine selection efficiencies and rates at each Trigger Level.•Apply isolation, energy, tracking matching etc cuts.
Trigger Level
% Efficiency w.r.t LVL1
Rate
LVL1 100 7.9 kHz
LVL2Calo 95.7 1.9 kHz
LVL2ID 88.4 359.1 Hz
LV2IDCalo 86.3 136.1 Hz
EFCalo 84.8 92.3 Hz
EFID 80.0 64.1 Hz
EFIDCalo 73.4 31.1 Hz
•Electron selection •Weνe sample
22.5 GeV
RAL Summer SchoolSeptember, 2004
Efstathios (Stathis) [email protected] http://www.hep.ucl.ac.uk/~sstef
Performance Studies: RoI size.How is the RoI constructed?
• Take the information from LVL1 Calo : (η,φ) of the active region. The η coordinate is calculated w.r.t to z=0
• Construct the shape of the RoI:
z=0 ± 168 mm (3 of the beam spread)
η ± 0.1
φ ± 0.1 radMotivation?
Optimize the size of the RoI.
Less Space Points.
Less Combinatories.
Quicker Algorithms.
Improve efficiency.Implementation?
Use the shower position at the 1st and 2nd EM sampling.
RAL Summer SchoolSeptember, 2004
Efstathios (Stathis) [email protected] http://www.hep.ucl.ac.uk/~sstef
Performance Studies: RoI size.
Z vertex resolution Phi vertex resolution Theta vertex resolution
σ_low = 3.626 cm
σ_high = 3.593 cm
σ_low = 1.719 mrad
σ_high = 1.692 mrad
σ_low = 14.26 mrad
σ_high = 14.11 mrad
RAL Summer SchoolSeptember, 2004
Efstathios (Stathis) [email protected] http://www.hep.ucl.ac.uk/~sstef
Future Plans.
•Performance Studies
IDScan : Optimize the size the RoI
e/γ analysis : Improve, validate, optimize the cuts
•Physics Studies
Low Mass Higgs (mH<150 GeV)
Hγγ
Hbb
Study the full production/simulation/reconstruction chain.