CHEP03 San Diego, 24 – 28 March 2003 adele rimoldi University of Pavia & INFN, Italy The full detector simulation for the ATLAS experiment: status and outlook

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 2 Atlas Simulation Project  As a global and coherent task it is at its starting point Geant3 Geant4 Fluka integration Detector Response Simulation Montecarlo Generators Fast simulation  Plan under analysis & discussion between participants LCG as one additional partner in the game ?  Other experiments with same problems Possible sharing for common solutions ?

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 3 Atlas: Managing complexity in the LHC environment Complexity A detector out of scale with respect to any previous ever built Demanding environment People vs things The biggest collaboration ever gathered in HEP (>2000 Physicists) The most complete and challenging physics ever handled Detector simulation program : an ambitious task Detailed simulation in Geant3 In action since 10 years Detailed simulation in Geant4 Blooming now from subdetector-centered activities Need to accommodate also Testbeam studies Staged detector environment for early use Physics studies Fast/semi-fast simulation Optimization, shower parameterizations, FLUKA integration…

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 4 A Road to the Atlas Simulation  Milestones: Data Challenges DC0 (end 2001 tests of event productions Geant3 & Geant4) DC1(Phase I ->II) Geant3 based Validation samples (single particle, Et scans,Higgs) 740K ev Single-particle production 30 million ev Minimum-bias production 1.5 million ev QCD di-jet production 5.2 million ev Physics events requested by HLT groups 4.4 million ev Pile-up New data samples requests from different communities DC2 and following large scale physics analysis, tests on computing model, test calibrations and alignment procedures Geant4 based (beginning 2004) Goal 10**7 fully simulated events Physics Validation Since 2001 Mainly for testbeam simulations and simple setups with updated tools

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 5 Simulation data flow GeneratorMcTruth(Gen)HepMC ROD Emulation Algorithm L1 Digitization Particle FilterSimulation PileUp McTruth(Sim)Hits ROD Input Digits McTruth(PileUp) DigitizationRawDataObjects ByteStream ConversionSvc MergedHits L1Digits L2Result EFResult L1 Emulation (inc. L1 ROD) L1Result ROD Emulation (passthru) L2 Selection Algorithm EF Selection Algorithm ByteStream Uses RawDataObjects ATLAS

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 6 Atlas & Geant3 package frozen since 1995 ATLAS still using Geant3 for the current Data Challenges Main parameters: 27 millions distinct volume copies 23 thousands different volume objects Processing time per job: about 24 hours Typical output file size for 170 – 320 ev (hits & digits): 200 – 300 Mbytes One, big blob…

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 7 Atlas & Geant4  Geant4 as a detector simulation toolkit of choice for Atlas currently most applications in vs now migrating to 5.0  2003: moving to Geant4 as the main simulation engine detector components ready and being assembled together Shifting emphasis from physics to computing after two years of physics validations (see Atlas talk on Physics validation issues – A.Dell’Acqua) To be addressed or already tackled: Memory usage minimization Performance optimization Initialization time monitoring/minimization Calorimeter parameterisation (crucial)

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 8 Atlas & Geant4: Framework  One production framework (Athena-Gaudi) which provides: basic services (persistency, condition DB etc.) Interface to event generators (via HEPMC) Event loop  But… A simulation framework (FADS/Goofy) On redhat7.2 gcc2.96 since 1 Year now on migration to redhat7.3.1 & gcc3.2 Most detector applications in FADS Most testbeam simulations are in FADS Advantages Faster Dynamic loading Actions on demand (uploading, downloading) No recompiling for most actions Modularity (see A. Dell’Acqua, A.Rimoldi)  There is still room for stand-alone applications Lar, muons, testbeams...

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 9 The Atlas detector in Geant4  Four main subdetectors Inner detector - momentum measurement of charged particles high precision silicon trackers straw tracker with TR capability Calorimeters - energies carried by particles measurements Lar EM and Endcap calorimeters Scintillating Tile calorimeter Muon spectrometer - muon identification and measurement High precision Drift Tubes for tracking, RPCs and TGCs for triggering Magnet system - bending of charged particles for momentum measurements

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 10 The Atlas Testbeams  One for each main subdetector at least Hot activity (see Physics Validation Programme >200 talks in 2 years) Precise detector response implemented carefully Always check/tune with data Special requests from simulation to data taking for special setups Al, Fe in front as in the MuonTestbeam 2002 case  Start first studies with combined setups Muon/tile/pixel First example of integration First need for using the same tools Mimic of a complete Atlas sector

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 11 The LAr Calorimeter Simulation  3 subsystems Electromagnetic barrel (EMB) Sampling calorimeter with LAr gaps electrode plates with accordion shape Hadronic end-cap (HEC) Forward calorimeter (FCal)  Heavy tests/investigations for optimizing the physics in Geant4 New multiple scattering model (different from Geant3) on the grounds of the Geant3 physics Energy loss process with ionization and bremmstrahlung not independent as in Geant3  Heavy tests for optimizing the geometry for reducing memory usage Voxelization smartless ~35 MB for testbeam setup EMB down to ~ 10MB, correspondent to an increase in time ~5%  Time in Geant4=1.7 Geant3

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 12 The Tile Calorimeter Simulation  Volumes Central barrel calorimeter module ~8300 volumes Extended barrel module ~4000 Total TileCal ~ Total testbeam ~25000  Execution time (PIII 800MHz Ram: 1GB) -> 36 sec/evt -> 14.4 ssec/evt -> 0.3 sec/evt GeV -> 73 sec/evt -> 29 sec/evt -> 0.4 sec/evt -> 100 sec/evt  Memory usage ~50 MB

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 13 The Muon System Simulation  A demanding geometry environment for a complex detector Four main subdetectors 2 precision chambers tracking detectors 2 trigger chambers detectors Many implementation tests Rewriting of code for accomodate new geometries in the detector design Setup changes – now on floor  The same program used for testbeam simulation Advantages of database vs plain numbers  Precise detector response in place for 2/4 detectors Hit collections different from digit collections  Nice connection to all muon reconstruction programs Simulation as a service for a better tuning of reconstruction programs  Physics studies already tackled successfully with G4 for A/H->2   Integration with toroids successfully done very recently  Testbeam simulation/reconstruction on floor also for combined setup Incident particle Testbeam Size 25 MB Muon System with Mag Field Size 73 MB (without 50) muonPIV 1600 MHz 256 MB Time/ev (sec) PIII 1266 MHz 512 MB Time/ev (sec) PIV 1600 MHz 256 MB Time/ev (sec) 10 GeV GeV GeV

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 14 Precision Chambers (MDT) Encap Trigger Chambers (TGC) Barrel Trigger Chambers (RPC) Inner Precision Chambers (CSC)

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 15

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 16 Atlas & Fluka  FLUKA as an integrated detector simulation program for background calculations radiation studies beam line simulations  Plans to integrate it: to be usable as a simulation engine for precise detector simulation studies Via FLUGG interface LCG project just started Needs to watch closely to its development

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 17 Atlas & shower parameterisation  Intermediate simulation level Faster than full simulation Integrated in the same framework To provide the same kind of output as full simulation (e.g. hits on which the full reconstruction chain can run) Encouraging results on a simple setup Longitudinal and later shower profiles well reproduced while execution time 100 times faster Now proceeding to parameterise EMB, TileCal etc…

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 18 Detector description matters  Currently working out of hard-coded geometry packages Basically from physics validations/detector studies Built using the same geometrical constants as G3 (via MySQL DB) No geometrical model passed to reconstruction  Already moving to a new system for DD/geometrical representation (GeoModel) Model built in memory using G4-like primitives G4 geometry built using (quasi-)automatic builders Possible interface to XML/GDML  Planning progressive replacement of existing sub- detector geometries as of this summer

25/03/2003Adele Rimoldi - University of Pavia & INFN, Italy 19 Some conclusions  We are on our way to completing the transition to Geant4 based simulations We are now working at a global plan for a coherent approach and comparisons between different simulation engines  The shower parameterisation approach is being actively pursued, in order to meet the increasing requirements on the detector simulation programs  Optimization and centralization of the detector description service is currently being tackled by moving to GeoModel and corresponding facilities Use case is the Muon system  Full geometry integration and optimization is now the main goal for meeting the deadline of next software release (7.0.0) and for making Geant4 the main engine for the Atlas detector simulation