Ties Behnke: Simulation and Tools 1 Simulation and Tools Ties Behnke, DESY Summary of the simulation sessions At this conference: 4 sessions (one combined with Loop Verein) Shows healthy interest in the subject Shows healthy number of people working with simulation tools This talk: as requested, no traditional summary but rather a view of the field and a (somewhat personal) perspective Areas covered in the session: Simulation proper Reconstruction algorithms and tools Generators (see separate generator summary)

Ties Behnke: Simulation and Tools 2 A Survey: where are we Legacy software LEP HERA... New developments JAS LCD GEANT3 GEANT4 BRAHMS MOKKA, Jupiter LCDG4,... The Linear Collider Community: experimentalists theorists software “nerds” PAW addicts..... C++ FORTRAN JAVA very few work full time on this! SNARK REPLIC VERTIGO PatRec

Ties Behnke: Simulation and Tools 3 Requirements Software for LC studies needs to be simple portable cheap (no commercial packages, if at all possible) rely on few number of well defined, easily accessible packages but at the same time provide full functionality... ??? make involvement of community simple remain open, close as few doors as possible (language...) keep it simple work together probably no “grand solutions” (they tend to be complex) concentrate on well designed simple interfaces don't be religious about anything

Ties Behnke: Simulation and Tools 4 Full Simulation GEANT4 accepted as de-facto standard GEANT3 still widely used (and still very powerful in many respects) But: the transition to GEANT4 is on-going validation of GEANT4 is happening The user community is growing GEANT4 is a toolkit (contrary to GEANT3, which was a “way-of-life”) many open auxiliary questions: geometry definition persistency graphics user interaction

Ties Behnke: Simulation and Tools 5 Fast Simulation Standard fast parametrised simulations (SIMDET, QuickSim) have been very powerful A move towards non-parametrised, but still fast MC is visible: allow detector optimisation allow fast changes of parameters (at least some) Lelaps (C++), SGV (Fortran) exist and work and are being used An event simulated in LeLaps: hits in tracker hits in calorimeter

Ties Behnke: Simulation and Tools 6 (Full) Simulation Frameworks MOKKA: developed since 99 geometry system (mysql database) infrastructure CGA (common geometry access) model LCDG4: developed since 03 (based on earlier version) xml based geometry model (simplified geometry at the moment) infrastructure persistency model Jupiter: uses root as frame and for persistency, GUI geometries in C++ (ParameterList classes to isolate geometry) BRAHMS: GEANT3 based full simulation and reconstruction package: still the most complete package available

Ties Behnke: Simulation and Tools 7 Commonalities while the frames are different the problems are the same! definition of geometries (accessible, sharable, transportable) definition of data model beyond pure simulation many detailed technical questions (e.g., MC tree treatment, etc etc) Lots of room for collaboration Lots of NEED for collaboration compared to the situation one year ago, there is significant progress in the migration towards GEANT4: encouraging! Plan: formulate a document with requirements and requested features for a full MC frame, hopefully with inter-regional agreement

Ties Behnke: Simulation and Tools 8 Architecture proposed architecture of a simulation / reconstruction system Generato r geometry Analysis Recon- struction Simulati on Java, C++, Fortran Geant3, Geant4 Java, C++, Fortran Data model / persistency

Ties Behnke: Simulation and Tools 9 Data model/ Persistency: LCIO LCIO (Linear Collider Input Output): proposal for a common data model and persistency scheme developed in collaboration between SLAC/ DESY/ LLR design goals: simple, lightweight package data model for LC studies, expandable to adopt to evolving needs supports broad set of platforms and environments support for C++ - JAVA - FORTRAN hit model reconstructed quantities model abstract “particle” model project started: Prague 2002 first beta release: Amsterdam 2003 first production release: end 2003 first complete release: LCWS 2004 project started: Prague 2002 first beta release: Amsterdam 2003 first production release: end 2003 first complete release: LCWS 2004

Ties Behnke: Simulation and Tools 10 Status Beta release of first complete implementation (including reconstructed quantities) available Many authors of LC packages are adopting LCIO: MOKKA (LCIO compliant) LCDG4 (soon to be LCIO compliant / merging with MOKKA) JUPITER (expressed plans to become LCIO compliant) LeLaps (LCIO compliant) BRAHMS (LCIO compliant) SIMDET (plans announced to be LCIO compliant) LCIO developments / status:

Ties Behnke: Simulation and Tools 11 Generators Generators are crucial (of course) for a successful study Flexibility is essential: legacy generators, modern generators, very modern generators.... sensible proposal: use stdhep as common interface between generators and simulation, support ascii and binary format large activity in the generator community: new complete packages reengineering of known packages (e.g. HERWIG++,...) more and more many fermion final state MC available from a user perspective: it would be nice to agree on a generic way to handle beam conditions (e.g. crossing angles,...)

Ties Behnke: Simulation and Tools 12 Reconstruction Lots of work on individual reconstruction tools is ongoing Tracking studies/ pattern recognition developments Algorithms developments for ParticleFlow merging algorithms clustering algorithms particle ID algorithms vertexing tools detailed detector modeling (noise etc)..... Currently no coherent frame exists for these developments study of the dependence of E-resolution on noise in the cal test beam and prototype efforts are really pushing the development! test beam and prototype efforts are really pushing the development!

Ties Behnke: Simulation and Tools 13 Reconstruction Example Example: particle flow / calorimeter reconstruction many groups do similar studies NIU Oxford UTA DESY ITEP.... (apologies if I forgot you) Sharing and comparing of results is very important and becomes much simpler if we speak a common language.

Ties Behnke: Simulation and Tools 14 Reconstruction: what's there RERECO: full reconstruction (tracking, calorimeter, particle flow) in FORTRAN LCIO compliant (at the moment the only complete reconstruction available) Several partial packages (clustering, particle flow, hybrid Monte Carlo reconstruction, LCD analysis framework (JAVA), root based analysis frames,...) The Goal: provide a simple framework for reconstruction and analysis modules to plug into provide interfaces to standard packages like WIRED (event Display) etc. The language of the reconstruction code should not be a central issue The Dream: make samples of fully reconstructed events available to the international user community through a “GRID” like framework

Ties Behnke: Simulation and Tools 15 Reconstruction Goal: integrate reconstruction tools as much as possible into the LCIO data model minimize dependence on external packages adopt the data model (which may need to be extended) advantages: ease of exchanging of tools (possibility of comparisons!) become more flexible save time and effort in the medium term In my view the advantages of a common API are far larger than the disadvantages of a “simplistic” system like LCIO

Ties Behnke: Simulation and Tools 16 The Future The collaborative model used for LCIO worked very well We should continue to develop this for other tools: Simulation Reconstruction Analysis Tools At this meeting: discussed and started even closer coherent development of GEANT4 frameworks Communication is very important: common WEB based Forum to exchange news, ideas etc We encourage everyone to use this forum. New forums can be added if required. This can be a very powerful tool! common WEB based Forum to exchange news, ideas etc We encourage everyone to use this forum. New forums can be added if required. This can be a very powerful tool!

Ties Behnke: Simulation and Tools 17 Conclusion Significant progress during the last year: LCIO is being adopted by more and more groups GEANT4 frameworks move to be real frameworks The user community is growing There is lots of progress in developing reconstruction algorithms There are still significant problems: Geometry definition is still an issue modern reconstruction frameworks do not exist personpower (very few developers for general tools) does hardly exist WE NEED YOU AND YOUR KNOW HOW TO MAKE PROGRESS AND SUPPLY THE TOOLS NEEDED BY THE COMMUNITY! WE NEED YOU AND YOUR KNOW HOW TO MAKE PROGRESS AND SUPPLY THE TOOLS NEEDED BY THE COMMUNITY!