L’ uso di GEANT4 negli Esperimenti LHC

1. L’ uso di GEANT4 negli Esperimenti LHC A. Rimoldi, Universita’ di Pavia & INFN adele.rimoldi@pv.infn.it CSN1, Perugia 11 Novembre 2002

2. Prodromo • L’utente/sviluppatore della simulazione ad LHC chi e’ quale ambiente ha trovato • Geant3 (G3) eGeant4(G4) Quali sono le sue prospettive nell’immediato futuro? • Qual e’ l’utenza (%) rispetto a quella legata all’area della ricostruzione e dell’analisi dei dati ? • Quale percentuale del suo tempo dedica in media alla simulazione ? • Qual e’ il suo rapporto con le simulazioni degli altri esperimenti (LHC)? Adele Rimoldi - Università di Pavia & INFN

3. La simulazione negli esperimenti e l’ LCG Da: priority issues for 2002 –LCG Launch Week Marzo 2002 (I.V.) • experiments have to participatein thedecision-making during execution of the project • NOW it is a good time to launch and implement common solutions • Experiments are in a period of important changes • integrating Geant4 & Fluka(Alice) into their frameworks • Full deployment of Geant4 • For CMS this means retirement of CMSIM during 2002 • For ATLASand LHCb validation of physics processes implies that Geant3 based simulations have still to be maintained for some time (2003) • Alice will reconsider use of G4 if physics improves. Adele Rimoldi - Università di Pavia & INFN

4. LHC: quale simulazione – Geant3 • Tutte le collaborazioni • stanno usando G3 • concordano nell’affermare che G3 verra’ dismesso • CMS & LHCb molto avanzate nella loro migrazione a G4 • le prossime DC in G4 • ATLAS sta avendo i propri DCs con G3 e sviluppando G4 • Ma DC0 gia’ effettuata (2001)con G4 (spettrometro a m+Tile) • Nel solo spettrometro a m in G4 10**7 ev gia’ prodotti per diversi studi • Accettanza geometrica • Muoni singoli e non, per tuning del programma di ricostruzione • tests di robustness • Alice e’ ancora basata su G3 ma esiste una interfaccia operazionale a G4 (via AliRoot) • Qual e’ la dimensione del ‘problema’, ad esempio in Atlas? • P.Nevski(ATLAS) talk (Software Week, March 2002), DC1: • 25,5 millions distinct volume copies • 23 thousands different volume objects • Processing time per job is about 24 hours • Typical output file size for 170 – 320 ev (h & d) is 200 – 300 Mbytes Adele Rimoldi - Università di Pavia & INFN

5. LHC quale simulazione – Geant4 • Tutte le collaborazioni • sono concordi in maggioranza nella scelta di G4 • Si sono impegnate nel programma di passaggio a G4 • richiedono che G4 sia piu’ (specialmente) focalizzato ai bisogni di LHC • richiedono che il programma di fisica sia piu’ indirizzato ai bisogni di LHC • soffrono per insufficienza di modularita’ o mancanza di funzionalita’ interattive • Cura attuata: uso di home-grown facilities (Iguana,Panoramix,Root) • cercano di adattare il G4Framework ai propri desiderata smontandolo e ad usarlo come un reale toolkit • Event generation, user actions, physics lists Adele Rimoldi - Università di Pavia & INFN

6. E Fluka? • Tutte le collaborazioniconcordano nell’affermare che: • E’ interessante poterlo forse considerare come un’alternativa a(lla fisica adronica di) G4 ma.. • Fluka dovrebbe essere supportato, mantenuto, documentato ed in generale pubblicamente disponibile • Sarebbe interessante e desiderabile avere un ambiente comune di simulazione che consenta di usare sia G4 che Fluka come ‘simulation engine’ partendo dalla stessa geometria, dallo stesso ‘event generator’ dalle stesse ‘user actions’ Adele Rimoldi - Università di Pavia & INFN

7. Una anticipazione • Dopo due anni di studi di ‘validazione della fisica’ (ad alta priorita’ in ATLAS, a bassa priorita’ in CMS) e in stretta collaborazione con il G4 team: • Il punto emerso da questo programma di lavoro e’ che • A livello macroscopico non ci sono problemi legati a G4 • In generalela fisica di G4 e’ in costante miglioramento e la fisica elettromagnetica e’ sicuramente migliore oggi in G4 • Ora c’e’ solo spazio per migliorare • studi di validazione • potenziamento delle human resources Adele Rimoldi - Università di Pavia & INFN

8. Geant4 perche’? • Ancora giovane ma con potenzialita’ di sviluppo • Parco utenti ancora modesto se non sotto soglia • C’e’ chi dice: ma non funziona a dovere, ancora • Ma occorre impegnarsi per farlo funzionare ampliando massivamente la comunita’ degli utenti (G3 docet) • Occorrono modifiche, anche di rotta La risposta di G4 talvolta non e’ stata la migliore Ma esiste un gruppo che ci sta lavorando aperto al dialogo Promuovere gli utenti degli esperimenti a sviluppatori (?) • I corsi di G4 (in Atlas): troppo presto? Troppo tardi… Adele Rimoldi - Università di Pavia & INFN

9. Sommario • I rivelatori di LHC e la simulazione: Frameworks Geometria Funzionalita’ In sintesi • Studi di validazione della fisica Il punto di vista degli esperimenti e di Atlas in particolare • G3 e G4 a un primo confronto quantitativo • Problemi • Conclusioni Adele Rimoldi - Università di Pavia & INFN

10. I rivelatori di LHC e la Simulazione in GEANT4 Adele Rimoldi - Università di Pavia & INFN

11. (Simulation) Framework • Tutte le collaborazioni hanno implementato propri frameworks per adattare meglio Geant4 al proprio profilo • CMS :Mantis is the COBRA interface to Geant4, a toolkit for building or interfacing to detector geometries and sensitive volumes, generators, physics, magnetic fields, actions etc OSCAR(2) is the CMS Geant4 simulation based on Mantis • LHCB :GAUSS: to use GEANT4 with the GAUDI Framework • Atlas :Athena Atlas’ chosen software framework, based on Gaudi: it provides SW environment, services, access to persistency… GOOFY: Simulation framework Completely based on G4 • Alice :AliRoot framework C++: 400kLOC + 225kLOC (generated) + macros: 77kLOC • FORTRAN: 13kLOC (ALICE) + 914kLOC (external packages) • Maintained on Linux (any version!), HP-UX, DEC Unix, Solaris • Works also with Intel icc compiler Adele Rimoldi - Università di Pavia & INFN

12. CMS barrel detectors CMS muon system View of 180 Higgs event simulated in CMS Tracker detector Geometria – CMS P.Arce CERN 30/9/2002 • All CMS detectors • Also several Testbeams • Currently moving to geometry in XML database, common to Simulation/Reconstruction/Visualisation • All geometry converted from GEANT3 geometry • Building of XML geometry into GEANT4 ready • Already can be visualised with IGUANA • Detailed checking quite advanced • Through comparison of GEANT3 and GEANT4 geometry Adele Rimoldi - Università di Pavia & INFN

13. Electromagnetic Barrel Accordion Calorimeter • 10 GeV Electron Shower Geometria – Atlas A.Dell’Acqua 18/9/2002 RH • Inner Detector • LAr calos • TileCalos • Muon System • TestBeam setup implementati per tutti i sottorivelatori • TOROIDI assenti Adele Rimoldi - Università di Pavia & INFN

14. Geometria: Alice Adele Rimoldi - Università di Pavia & INFN

15. Funzionalita’ – LHCb S.Easo RAL 30/9/2002 • Geometry Input: XML database. A version available for all • the detectors in LHCb. • Input events: From Pythia or other similar programs • through the HEPMC interface into GEANT4. • A first version of the whole Simulation chain is now working. • Starting to study the response of the detectors in detail. Adele Rimoldi - Università di Pavia & INFN

16. Funzionalita’ - CMS P.Arce CERN 30/9/2002 Physics: CMS Physics Technical Design Report is planned to be written using OSCAR (due end 2005) OSCAR milestone Spring/Summer 2003: “Reproduce the physics results of the GEANT3 simulation with similar performance” essential for DC 2004 and Computing TDR at end 2004 UserActions: • Several user actions of the same type loadable on demand Documentation: • User’s Guide updated for each release • Several Tutorials on the web • For OSCAR developers • For Physics Reconstruction software developers • For Summer Students OSCAR is ready for physics validation studies Adele Rimoldi - Università di Pavia & INFN

17. Funzionalita’ - Atlas A.Dell’Acqua– RH 18/9/2002 • hits (digits soon) passed to Athena automatically (via G4Svc) can use both the Athena generators and the Goofy ones detectors can be added/removed dynamically improved user action system support for XML, MySQL, etc. ready for “release” A.Dell’Acqua: Software Week Ottobre 2002 • We are right in the middle of a big, “Big Bang” style move to Geant4 as the main simulation engine • The detector will come together at once • Shifting emphasis from physics to computing for a while • Are we in for some nasty surprise? • Memory, performance, initialization time etc. ARE a concern • We are working hard on parameterising our calorimeters Adele Rimoldi - Università di Pavia & INFN

18. In sintesi Adele Rimoldi - Università di Pavia & INFN

19. La ‘Time performance’: G3 vs. G4 • Atlas: • All’inizio della fase di checking ma… “Atlas FCAL: Geant4 is 3x faster than Geant3” “Atlas EM Barrel (2001): comparable performance” • CMS: • - Full events: H / tte / Z • - Cuts in primary particles: Pt : 1 GeV, || < 2.4 (3.0) - Full CMS geometry - 3D TOSCA magnetic field - Production and tracking cuts as in CMSIM (GEANT3) - GEANT4 voxel navigation - GEANT4.4.0.ref02 - Pentium III 850 Mhz - Checked that they are really the same events (see next) - H (10 events) CMSIM: 439k tracks 64 sec/evt OSCAR: 400k tracks 117 sec/evt (1.83) - tte (10 events) CMSIM: 1809k tracks 184 sec/evt OSCAR: 1159k tracks 425 sec/evt (2.31) - Z (10 events) CMSIM: 1105k tracks 96 sec/evt OSCAR: 941k tracks 248 sec/evt (2.58) Non ottimizzazione! Adele Rimoldi - Università di Pavia & INFN

20. Le validazioni della fisica in G4 • CMS si trova in un programma di simulazioni per validazioni di fisica piu’ approfondito che in passato • LHCb, sta tunando la propria simulazione sulla base dei risultati di testbeam • Atlas ha immesso energie da piu’ di due anni con un programma dedicato • Dal talk di F.Gianotti alla Software Week di Atlas dello scorso marzo: • G4 Physics validation in ATLAS started in June 2000: • -- organised by Simulation Coordinator and Physics Coordinator • -- software, detector and performance people involved • -- very fruitful collaboration with G4 experts • -- monthly meetings attended also by G4 experts + 2-day workshop fall 2001 • -- > 150 presentations on physics validation given by ATLAS or G4 experts Adele Rimoldi - Università di Pavia & INFN

21. Quali risultati – Atlas P.Loch ultima Atlas Week Ottobre • Geant4 physics benchmarking: • compare features of interaction models with similar features in the old Geant3.21 baseline (includes variables not accessible in the experiment); • try to understand differences in applied models, like the effect of cuts on simulation parameters in the different variable space (range cut vs energy threshold…); • Validation: • use available experimental reference data from testbeams for various sub-detectors and particle types to determine prediction power of models in Geant4 (and Geant3); • use different sensitivities of sub-detectors (energy loss, track multiplici-ties, shower shapes…) to estimate Geant4 performance; • tune Geant4 models (“physics lists”) and parameters (range cut) for optimal representation of the experimental detector signal with ALL relevant aspects; Adele Rimoldi - Università di Pavia & INFN

22. La validazione della fisica - Atlas • Hadronic EndCap Calorimeter (HEC) • (Liquid Argon/Copper Parallel Plate) 10-1 10-2 180 GeV μ Events/10 nA Fraction events/0.1 GeV 10-3 Eμ= 100 GeV, ημ ≈ 0.975 10-4 800 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 700 Reconstructed Energy [GeV] Calorimeter Signal [nA] 600 0 500 -0.5 • G4 simulations (+ electronic noise) describe testbeam signals well, also in Tile Calorimeter (iron/scintillator technology, TileCal); • some range cut dependence of G4 signal due to contribution from electromagnetic halo (δ-electrons); -1.0 400 -1.5 300 Δ events/0.1 GeV [%] -2.0 200 -2.5 100 -3.0 0 -3.5 400 -100 0 100 200 300 500 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Reconstructed Energy [GeV] • Electromagnetic Barrel Calorimeter • EMB (Liquid Argon/Lead Accordion) Muon Energy Loss Adele Rimoldi - Università di Pavia & INFN

23. 2 0 -2 -4 -6 2 0 -2 -4 GEANT4 GEANT4 -6 GEANT3 GEANT3 data data 0 1 2 3 4 5 GEANT4 GEANT4 GEANT3 GEANT3 0.2 0.3 0.4 0.5 9 9.2 9.4 9.6 La validazione della fisica - Atlas Geant4 Electron Response in ATLAS Calorimetry • Overall signal characteristics: • Geant4 reproduces the average electron signal as func- • tion of the incident energy in all ATLAS calorimeters • very well (testbeam setup or analysis induced non-line- • arities typically within ±1%)… • …but average signal • can be smaller than in G3 • and data (1-3% for 20- • 700 μm range cut in HEC); • signal fluctuations in EMB • very well simulated; • electromagnetic FCal: • high energy limit of reso- • lution function ~5% in G4, • ~ 4% in data and G3; FCal Electron Response EMB Electron Energy Resolution ΔErec MC-Data [%] • TileCal: stochastic term • 22%GeV1/2 G4/G3, 26%GeV1/2 • data; high energy limit very • comparable; Adele Rimoldi - Università di Pavia & INFN

24. La validazione della fisica - Atlas Geant4 Hadronic Signals in ATLAS Calorimeters • Calorimeter pion response: • after discovery of “mix-and-match” problem (transition from low energy to high energy char-ged pion models) in the deposited energy from energy loss of charged particles in pion showers in the HEC (G4 4.0, early 2002): fixes suggested by H.P. Wellisch (LHEP, new energy thresholds in model transition + code changes) and QGS model tested; • e/π signal ration in HEC and TileCal still not well reproduced by Geant4 QGS or LHEP - but better than with GCalor in Geant3.21; • energy dependence in HEC in QGS smoother, “discontinuities” between ~20 GeV and ~80 GeV gone; HEC Pions QGS e/π signal ratio LHEP e/π signal ratio Pion energy [GeV] Adele Rimoldi - Università di Pavia & INFN

25. Geant4 Hadronic Signal Characteristics (1) • Pion energy resolution: • good description of experimental pion energy resolution by QGS in TileCal; LHEP cannot describe stochastic term, but fits correct high energy limit; • larger discrepancies in HEC: stochastic term significantly too large for QGS and LHEP, but reasonable description of high energy behaviour by QGS; TileCal Pion Energy Resolution HEC Pion Energy Resolution Exp QGS GCalor relative energy resolution [%] Pion energy [GeV] La validazione della fisica - Atlas Adele Rimoldi - Università di Pavia & INFN

26. La validazione della fisica - Atlas • Geant4 can simulate relevant features of muon, electron and pion signals in various ATLAS detectors, often better than Geant3; • remaining discrepancies, especially for hadrons, are addressed and progress can be expected in the near future; • ATLAS has a huge amount of appropriate testbeam data for the calorimeters, inner detector modules, and the muon detectors to evaluate the Geant4 physics models in detail; • feedback loops to Geant4 team are for most systems established since quite some time; communication is not a problem; • lack ofman power in all ATLAS systems makes it hard to follow up on Geant4 progress; hard to catch up with G4 evolution! Need to look into more automated benchmark tests ?? • Geant4 is definitively a mature and useful product for large scale detector response simulations! Adele Rimoldi - Università di Pavia & INFN

27. Problemi dalla visuale di Atlas • (wo)manpower : assolutamente critico • Mancanza di risorse, dalla parte dei rivelatori • Alcune aree dei detector non sono ancora coperte o hanno bisogno di una rivisitazione sostanziale • Anche dal punto di vista dell’ottimizzazione delle performances • Clashes con altre attivita’/requirements • esempio: Detector Description (Atlas,Muoni) • C’ e’ un “dependency problem” tra i vari domains ! • Alcune aree che logicamente o storicamente appartengono alla simulazione non sono state ancora prese in considerazione o vanno rivisitate • Pile-up, digitizzazione • Documentazione chi la scrive? Adele Rimoldi - Università di Pavia & INFN

28. Conclusioni • Le attivita’ relative alla simulazione in G4 sono decollate nelle collaborazioni • Occorre fornire un servizio migliore • Occorre (continuare a) dare un contributo fattivo alle validazioni di fisica • Occorre aumentare le human resources a breve termine • per avere massa critica • per trovare una soluzione rapida ai problemi Adele Rimoldi - Università di Pavia & INFN

29. Extra slides Geant3 Geant4

30. Geant3 Submitted: 01.10.84 Revised: 20.04.94 • GEANT is a system of detector description and simulation tools that help physicists in such studies. The GEANT system can be obtained from CERN as six Patchy /CMZ files: GEANT, GEANG, GEANH, GEANF, GEANE and GEANX. The program runs everywhere the CERN Program Library has been installed • The GEANT and GEANG files contain most of the basic code. The GEANH file contains the code for the hadronic showers simulation from the program GHEISHA []. The GEANF file contains the source of the routines for hadronic showers development from the FLUKA program which is interfaced with GEANT as an alternativeto GHEISHA to simulate hadronic cascades. The GEANE file contains a tracking package to be used, in the context of event reconstruction, for trajectory estimation and error propagation. The GEANX file contains the main program for the interactive version of GEANT (GXINT) and a few examples of application programs which may help users to get started with GEANT. Adele Rimoldi - Università di Pavia & INFN

31. Geant4 • Geant4 provides a complete set of tools for all the domains of detector simulation: Geometry, Tracking, Detector Response, Run, Event and Track management, Visualisation and User Interface. An abundant set of Physics Processes handle the diverse interactions of particles with matter across a wide energy range, as required by Geant4 multi-disciplinary nature; for many physics processes a choice of different models is available. In addition a large set of utilities, including a powerful set of random number generators, physics units and constants, Particle Data Group compliant Particle management, as well as interfaces to event generators and to object persistency solutions, complete the toolkit. • Geant4 exploits advanced Software Engineering techniques and Object Oriented technology to achieve the transparency of the physics implementation and hence provide the possibility of validating the physics results. User Requirements have been collected in the initial phase of the project and are periodically updated. • Problem domain decomposition and Object Oriented Analysis and Design have led to a clear hierarchical structure of sub-domains, linked by a uni-directional flow of dependencies. The Geant4 Object Oriented design allows the user to understand, customise or extend the toolkit in all the domains. At the same time, the modular architecture of Geant4 allows the user to load and use only the components needed. • Since 1999 the Production Service, User Support and development of Geant4 have been managed by the international Geant4 Collaboration, which is based on a Memorandum of Understanding among the participating Laboratories, Experiments and National Institutes. Many specialised working groups are responsible for the various domains of the toolkit. • The Geant4source code is freely available, accompanied by an Installation Guide and an extensive set of documentation. Adele Rimoldi - Università di Pavia & INFN