Simulation Project Leaders Meeting 11-January-2011 ==== Attendees John Apostolakis (JA), Alfredo Ferrari (AF), Andrea Dotti (AD), Witold Pokorski (WP), Pere Mato (PM), Gabriele Cosmo (GC). Excused: Rob Veenhof (RV) --------------------------------------------------------------------- ==== Minutes - New technical student, Anton Pytel, joined the Generator Services (will contribute to restructuring of the project's web pages, integration of the HepMC analysis tool tests in the suite, and to improve the web tools for the MCPLOT pages). - New fellow, Victor Gonzalez, started in December to work on SPI/Geant4 system testing. - Re-organisation of the SFT group: Simulation project (Generators, Geant4, Validation, G.Cosmo leader), Root project (F.Rademakers leader), SPI (B.Hegner leader). New R&D activity in detector simulation by R.Brun, with resources mainly outside of the SFT group. J.Harvey is now appointed as the new SFT group leader; P.Mato is leading the support and technical coordination. --------------------------------------------------------------------- ==== Simulation physics validation & framework - Andrea - Meetings: o The last LCG Physics Validation meeting has been held on December 1st and was mainly dedicated to validation of the Geant4 9.4 release (from CMS) and to report on the developments ongoing to address ALICE's request for the simulation of anti-p, anti-d, anti-t and anti-He. o The presentation from CMS showed the comparison of 9.4 release with CMS test-beam data. It has been shown a good agreement with data for the "production" physics lists: QGSP_BERT_EML, FTFP_BERT_EML, QGSP_FTFP_BERT_EML (which replaces the parameterised model with FTF). Comparison with reference-tags timely made in CMS, has shown no large deviation in the predictions from the latest versions. Anti-proton and kaon responses are now routinely checked. This last study is in line with what was observed up to now: the most promising physics list is QGSP_FTFP_BERT (and its variants). o Two presentations from Geant4 developers have shown progress reports on the implementation of anti-p and anti-ions simulation in Geant4. In particular, the treatment of cross-sections in the Glauber-Gribov approach with the Grichine parameterisations have been discussed. Good description of the total, elastic and inelastic cross-sections of these particles with matter has been reached comparing data with parameterisations and the original GG theory. This is the first fundamental ingredient to implement a detailed anti- ions simulation in Geant4. These new developments have been included in Geant4 9.4. o Finally a presentation form ATLAS Hadronic Tile Calorimeter has shown a study to address the systematic on absolute energy calibration of the calorimeter using muons. This study rely on the correct simulation of these particles from MC that is used to quantify the systematic uncertainty of the e.m. scale. The study concluded with an absolute systematic uncertainty of about 1.2% on the ratio (from MC) of e/mu. Most important the study has shown an unexpected drop of the response, for the TileCal setup, on the response at low production cut values for electrons and gammas. o Next Meeting is scheduled for (end of) February (to be confirmed). - Validation Studies: o The EM group has started the study of the dependence of the response on cut value for the TileCal setup. The effect has been confirmed also on simplified setups. However the source of this effect is not yet clear. o Preliminary tests from CMS has shown some instabilities (crashes) for the 9.4 release in physics-lists making use of the CHIPS module (QGSP_FTFP_BERT_EML). These instabilities have not been observed during the release period in physics lists different from the CHIPS one, however from the log messages the problem has been identified and addressed by CHIPS author in a recent bug fix. The patch has been tested in Geant4 (waiting also for response from CMS) and will be included in a next patch together with other fixes. o A loss of performance of the order of 2-3% is experienced in 9.4 under some circumstances compared to 9.3.p02, it is still not clear the origin of this, under investigation. o Technical discussion on migrating Project static web-pages to SFT DRUPAL has started, the idea is also to add validation plots generated on the fly with the validation testing suite. --------------------------------------------------------------------- ==== Generator services - Witek - Meetings: o Meeting on November 10th, where the status of EvtGen was discussed. An agreement was reached to have now a unique group of people taking care for the development and maintenance of the package. - GENSER: o The porting on MacOSX (64 bits) has been completed, also with documentation. All new generators are now ported; if requested, builds can also be provided for old generators. o New versions of Pythya-8 and Tauola installed. - MCDB: o There have been some discussions about the way the service is run, which is currently provided by the developers; envisaged the possibility to provide implementation of Lemon and Quattuor sensors so that the status of the service can be monitored directly within IT; CMS may help in this with resources; to be discussed with the CMS management. --------------------------------------------------------------------- ==== Fluka - Alfredo - Meetings: o There has been a Collaboration meeting before Christmas, where first comparisons of DPMJet with data from LHC have been shown, and look promising. Improvements in the code to make the match even better are planned to be included in the next new release of Fluka. o Elections for the scientific committee have been taken, involving 5 members with expired mandate; some renewed and a new member elected, Vincenzo Patera from Frascati. - Developments: o The new release Fluka-2010.4 is expected for beginning of February. The new release will also include a new module of Flare for the graphical interfaces, now allowing to plot, manipulate/rotate geometries, debugging with also suggested corrections. o The pre-release, under testing since the last 4 months also includes a new BME version for low-energy ion interactions, which has now demonstrated to be pretty stable. The final tuning of the DPMJet code is in the pipeline before getting to the production release. o A lot of new development for heavy ion stopping power, requiring finalization. o Finalized all work for damage predictions (DPA non-ionizing energy loss prediction, etc..). --------------------------------------------------------------------- ==== Geant4 - John - Releases: o Geant4 9.4 was released on schedule. Among the features included, to mention: + A new tuning of the Urban multiple-scattering EM model and adoption of a dedicated model for multiple-scattering of muons, giving better physics (the old global model has been removed in this release); + New optional model for Bremmstrahlung angular distribution, mainly affecting lower energies; + Adopted improved modeling of hyperons and anti-nucleons (using CHIPS in place of LEP) in several emerging (promising) physics-lists: FTFP_BERT, QGSP_FTFP_BERT. As QGSP_BERT retains LEP for hyperons and anti-hyperons, a variant QGSP_BERT_CHIPS follows the new convention. + Removed many obsolete physics lists, especially old and transient configurations. This included many which used LEP models for energies below 5-15 GeV (FTFP, FTFP_EMV, FTFC, QGSC, QGSP_DIF, QGSP_EMV, QGSC_EMV, QGSP_EMX). Transient variants deleted included configurations without a separate quasi-elastic channel (NQE). and early/experimental QGSC variants QGSC_QGSC and QGSC_EFLOW. Also deleted were QGSP_QE, QGSP_BERT_DIF, and variants featuring the LHEP parameterized models: LHEP_BERT_HP, LHEP_BERT, LHEP_PRECO_HP. + New fast option (XS) for neutron cross-sections can improve CPU performance, in case of application requiring accuracy below 20 MeV. Based on interpolation of HP (G4NDL data library) cross-sections; + Significant improvement of the Fritiof/FTF modeling at low energy (3-8 GeV), using Reggeon cascade, better selection of the final state using phase-space; + Implementation of Bertini cascade significantly reviewed and improved, eliminating memory churn (negligible allocation) and improving conservation of energy/momentum. Revised internal pion cross-sections. Added new optional interface with pre-compound de-excitation model (all developments carried by the SLAC-Geant4 team - SFT team collaborated on validation and testing); + Improved CHIPS cross-sections: moved to isotope-wise (not element) cross sections for neutrons; new anti-baryon on nucleon cross sections; + Improvements in the CHIPS elastic process: added anti-baryon-nucleon elastic scattering; [* Note: many CHIPS improvements were delivered in 9.4-beta. During July-December issues were identified in tests and validation; fixes and corrections were found and smaller improvements were made.] + Improvement enabled the use in production of Fermi-breakup for light nuclei and evaporation of heavier fragments (Univ. Sevilla); relativistic improvements in JQMD ion-ion (SLAC); + In geometry, new G4GenericTrap solid (ALICE request); first version of precise safety algorithms; possibility to introduce gaps in divisions; + First implementation of a new build/installation environment based on CMake. All details and much more can be found in the release notes: http://cern.ch/geant4/support/ReleaseNotes4.9.4.html - Ongoing developments: o Progress on modeling for anti-ion interactions in the FTF model. o Planning a revisit of the QGS code once the extenstion/tuning of FTF is completed; o Preparing to extend existing interface to DPMJet-2.5 to nucleon projectiles. o Interest in utilising also DMPJet-3; would benefit from improved visibility for this generator - JA and AF will investigate. - Infrastructure: o The move to SVN of Geant4 source repository is going on these days. An automatic mechanism has been put in place in order to migrate the whole repository from CVS, keeping tags and histories. - Validation: o An issue related to CHIPS was found after releasing 9.4. A fix has also been provided to CMS and will be included in a next patch. - Feedback from experiments: o CMS: first pre-production tests on 9.4 found extra time cost, apparently mostly due to the new physics-list setup used (QGSP_FTFP_BERT_EML vs QGSP_BERT_EMV). o ATLAS: requested to have the EM variants available. The problem on stuck tracks reported some time ago is under active investigation. First results find that it is not frequent. Performance is reported unchanged in first tests on release 9.4. Extra memory cost reported (~70 MB). - Meetings: o Input from the last TF of December: asked to move forward the date of 2011 release (CMS, LHCb). A shift in the EM fraction from 9.1 and 9.2.p04 was reported by LHCb - it has not been confirmed though that the physics-list chosen was the same. --------------------------------------------------------------------- ==== Garfield - Rob - No report. --------------------------------------------------------------------- ==== AOB Next meeting scheduled for March 1st.