Simulation Project Leaders Meeting 30-April-2013 ==== Attendees (virtual meeting) John Apostolakis (JA), Alfredo Ferrari (AF), Witold Pokorski (WP), Rob Veenhof (RV), Gabriele Cosmo (GC). --------------------------------------------------------------------- ==== Minutes --------------------------------------------------------------------- ==== Simulation physics validation & framework - Witek, Alberto - Activities: o Validation studies made on the recent development releases of Geant4 reveal some differences in QGSP_BERT and FTFP_BERT between Iron and Copper, and between Tungsten and Lead; it has been identified that most of this is due to recent changes introduced in the development version of the Bertini code, which is now under investigation. o A new tool for model-level testing of calorimeter energy response based on PyRoot has been implemented, providing a bridge between comparison with thin-target experiments and full calorimeter tests. Key uses are for comparisons between models and regression testing, to check whether simple energy sums (per particle) have been changed when a model has been revised or fixed; for a given observable, it is now possible to compare different hadronic models (e.g. BERT, FTFP, BIC, QGSP) for the same or different Geant4 versions. - Meetings: o Overall satisfaction on Geant4 physics performance has been expressed at the recent CHEF-2013 conference held in Paris on April 22-25th by the LHC experiments and CALICE; still, improvements are requested related to the shower descriptions which remain too narrow. No feedback has come yet from CALICE on the new developments made on the physics models in the more recent releases of Geant4. It has been expressed the wish to have the CALICE setups included in the Simplified Calorimeter test-suite. This has been put now in the to-do list and will be implemented soon. o Contacts with experiments representatives have started in order to fix a suitable date for holding a second LPCC workshop on physics validation, following the first one held in October 2011. --------------------------------------------------------------------- ==== Generator services - Witek - GENSER: o The effort is currently being concentrated on restructuring of GENSER system and move it to CMake. More than half of the generators have been already migrated. The end of May has been fixed as deadline to complete this task and to integrate GENSER entirely with the LCG external software. o New generators installed: Herwig++ 2.6.3, pythia8.176, Rivet 1.8.3. --------------------------------------------------------------------- ==== Fluka - Alfredo - No report. --------------------------------------------------------------------- ==== Geant4 - John, Gabriele - Development: o Eliminated sampling in the conversion from physical step length to geometrical step in G4UrbanMscModel. In its place the size of the first step of a track and the first step in each volume is randomised in order to estimate back scattering, instead of sampling z (geometrical step length). This fixes a known issue of zero step length for ionisation, bremsstrahlung and annihilation in flight observed recently. A new geometrical limit of 0.01 nm has been added to allow scattering in very small steps. o The use of improved safety in voxelised parts of a geometry has been enabled; the value is no longer limited by the nearest boundary of a voxel. In the past the boundary of the equivalent voxel in which a point was located were seen as 'walls', beyond which the volumes were not searched. Changes in the value of the safety could lead to change in CPU usage: the better estimate costs extra cycles, but could reduce work during tracking in field. In future it could also reduce generation of secondaries when low safety was obtained despite being far from boundaries. o Hadron inelastic choices in physics lists have been revised; replaced the obsolete LEP with FTFP and BERT in all physics lists which had the parameterised LEP and/or HEP models. These include QGSP_BERT, QGSP_BIC, QGS_BIC, INCLXX and their derivate lists, such as QGSP_BERT_HP. After this revision, QGSP_BERT now differs from QGSP_FTFP_BERT only in the choice of transition regions: in QGSP_BERT the Bertini model is used more, up to the transition in the range 9.7-9.9 GeV to FTFP, whereas in QGSP_FTFP_BERT the transition is in the range 6-8 GeV. o G4MuPairProductionModel: extended grid of differential cross-section tables (from 8 to 40 points in muon energy). Use faster binary search of index in the table. Added check on maximum energy to reduce number of bins in muon energy. o Obsolete models, G4UrbanMscModel90, G4UrbanMscModel92 and G4PEEffectModel have been removed. o In materials, now always creating an isotope vector with natural abundances for each element, if one is not provided. o Introduced multi-body final state momentum distribution classes and factory, modelled on the two-body angular distribution system. Being used to address issues found in Bertini for the final states for five or more secondaries above 5 GeV. o The physics lists module has been restructured. The new structure includes submodules for hadronic elastic, hadronic inelastic, hadronic stopping, ions, decay and electromagnetic. o External packages have been consolidated into the directory 'externals'. o Added file compression for histograms in analysis module. - Multi-threading o Multi-threading build is now enabled in the latest monthly development release, with current limitation to a single run. o Created and started use of platform independent locking in new classes G4AutoLock and G4Threading. o Refined API for multi-threaded Run Manager o Reviewed several classes to identify constants which can be shared between threads. o Extended Scoring meshes to sum (reduce) results from different threads. Histograms of threads are summed automatically before writing in analysis package. o Significant cleanup of changes for multithreading in many classes. - Fixes in development version: o Fix in G4SteppingManager for proper handling of multiple AtRest processes. o Corrected an issue with the precise safety in voxelised geometry. o Corrected an issue in G4ExtrudedSolid for case affecting ATLAS Boolean subtraction of solids. o Identified and fixed issues in G4Tet in DistanceToOut() and SurfaceNormal() methods. o Corrected magnetic moment in mu+, other leptons and proton. o A correction for a problem with the time of flight of neutrons has been prepared. The problem was caused by the lack of an update of the velocity after elastic scattering. This resulted in an increasing deviation in the global time of flight with the number of consecutive elastic collisions between inelastic collisions (which did reset the velocity). - Experiments feedback: o ATLAS has evaluated a corrected version of multiple-scattering with fixes for the angular deflection in small steps in gas (order tens to a hundred microns) within its production that is based on Geant4 9.4.p04. Improved versions of Urban (Urban95), gamma conversion and Bremsstrahlung were ported to 9.4.p04 and tried out. Results are stable between Urban93 and Urban95, but agreement with data is not yet achieved. Also utilised corrected relativistic cross-section for GammaConversion and improved angular deflection in Bremsstrahlung, which did not provide a significant difference in shower observables. A discrepancy between MC11 and MC12 for the Z line shape was reduced but not completely eliminated by the MSc fix. A discrepancy in the lateral shower shape compared with collider (and test beam) data is not changed substantially by the revised Bremsstrahlung or multiple-scattering. o CMS has validated Geant4 9.6.p01 for production. Calorimeter observables agree with the previous production version. A discrepancy in tracking is under investigation. CMS awaits fixes for a number of issues, including neutron time of flight. o LHCb reported an issue found in the production with Geant4 9.5.p02, with an apparent infinite loop in one hadronic model. It seems that changing of the values of masses for resonances from the Geant4 default to newer values is triggering this. A proposed fix is under evaluation. No feedback yet on the use of the improved G4TesselatedSolid for the RF foil. o A first profiling of the ALICE simulation configure to use Geant4 was carried out, and the results were shared with the ALICE simulation team. - Meetings: o The development plan for 2013 was presented at the last Geant4 Technical Forum on March 26th. It is available from the Geant4 web site at: http://cern.ch/geant4/support/planned_features.shtml o The 9th Geant4 Space Users' Workshop took place in Barcelona on March 4-6th. o A Geant4 Course is being offered at the 10th Seminar for Nuclear, Sub-nuclear and Applied Physics, Porto Conte, Alghero (Italy), 2-8 June 2013. --------------------------------------------------------------------- ==== Garfield - Rob - Nothing to report. --------------------------------------------------------------------- ==== AOB Next meeting scheduled for July 9th.