The interface to Evt Gen in CMS Roberto
The interface to Evt. Gen in CMS Roberto Covarelli University of Rochester Evt. Gen workshop, 06/12/2010
Outline p p Introduction to the CMS framework The interface to Evt. Gen. LHC in CMS: n General structure p n n The Pythia-Evt. Gen generation chain Technical details and documentation Peculiar features of implementation: The usage of “alias” particles p The use of external generators (Pythia, PHOTOS) p p Some tests of Evt. Gen for inclusive productions Conclusions Remaining issues 2
The CMS framework (CMSSW) p Code compilation/linking (C++): n n Automatic generation of Make. File and compilation (scram) Library linking (in order): p p p n p User-defined libraries in local areas Standard CMSSW packages (a set of compatible package versions is altogether referred to as a release) “External” (non-CMSSW) libraries, e. g. MC generators Single executable application as output (cms. Run) Configuration and running (Python): n cms. Run driven by a configuration file p n It contains a schedule of modules to be run in the specified order Output information stored in ROOT file format 3
Evt. Gen. LHC in CMSSW p Event generation flow: 1) Run e. g. Pythia 6 as the event source p 2) Run Evt. Gen. LHC as an “external decay driver” p p p 3) Particle types known from Evt. Gen tables artificially made stable Decay “undecayed” particles that are in Evt. Gen tables Inclusive B decays (i. e. those whose BR’s are not specified) are generated via external interface to Pythia 6 Radiative corrections calculated via interface to PHOTOS Output stored as CMS Hep. MCProducts p Decay products are translated from standard HEP to Hep. MC format 4
Technical details/documentation p Technical details: n n Tested in CMSSW up to 3. 9. 0 releases External packages needed: Evt. Gen. LHC 9. 1 (officially maintained in GENSER) p PHOTOS 2. 15. 5 (officially maintained in GENSER) p § Upgrade of 2. 15. 3, required by ATLAS to inhibit FORTRAN “stop” statements § Required by CMS in order to fix mess with internal /HEPEVT/-like common blocks (see next slides) p Documentation: On TWiki: https: //twiki. cern. ch/twiki/bin/view/CMS/Evt. Gen. Interface n 5
“Alias” particles: implementation p A bit involved because three kinds of aliases are possible: 1) Aliases that are part of the decay tree of another Evt. Gen alias p 2) Alias particles originally produced by PYTHIA p 3) Treated normally by Evt. Gen If more than one in the event, one is randomly chosen to be the alias, others are decayed normally Aliases that are part of an Evt. Gen decay tree where the mother is NOT an alias (e. g. B 0* B 0 g, B 0 signal mode) p This is a particular case, since Evt. Gen generates the initial decay tree in one go: so daughters must be scanned and, if aliases are found, their products must be dropped and re-generated see timing No input required by the user (besides decay file) “Top-level” aliases: names must be provided by the user in a separate “forced_decay” string 6
External generators p PYTHIA: n n p PYTHIA called inside Evt. Gen itself to generate inclusive B decays, e. g. multi-body with unknown BRs and simple phase space distribution OK, but initialization forced within the interface to avoid overwriting of PYTHIA event source parameters (like stable status for B hadrons) some weird effects observed!!! PHOTOS: n n n Originally using /HEPEVT/ common block for C++/Fortran data exchanging Conflicts with CMSSW simulation-reco chain not understood Solved with: p p PHOTOS 2. 15. 5 (currently default in CMSSW) In the interface, with a new Evt. Gen-PHOTOS data exchange method which completely removes use of common blocks (overrides standard Evt. PHOTOS) 7
Evt. Gen for inclusive samples p p Is Evt. Gen suitable/better for inclusive sample generation? Test comparisons with PYTHIA 6: n n n p Timing Key variables in ttbar samples (used in b-tagging) Key variables in QCD samples (all Evt. Gen decays activated, including light mesons… etc. ) ongoing Evt. Gen generation adds 5. 5 ms CPU-time per event w. r. t. PYTHIA n n Observable in a standalone generation task (1. 5 h for 106 events) or in productions with high rejection in generator-level filters Dominated by: p p recursive search for alias particles light meson decays optimization? 8
ttbar samples (I) p Study by G. Giurgiu et al. : comparison of MC truth variables relevant for b-tagging n PYTHIA vs. Evt. Gen with long-lived particle decays inhibited 9
ttbar samples (II) p Good agreement found in most variables n n Evt. Gen predicts a slightly smaller number of tracks per B vertex Discrepancy observed in b e spectrum at very low values of p. Trel (lepton trasverse momentum relative to B hadron direction): under investigation 10
Attempt to use “merged” version Original version has compilation problems inside the CMS framework p No problem using the “merged” version as implemented in LHCb p n n No big changes required to use it in CMSSW (e. g. PHOTOS, see previous slides) Seemingly big difference in how CPV is implemented w. r. t. the GENSER version 11
Conclusions p The CMSSW interface to Evt. Gen is working fine: n Most delicate tasks to be dealt with: “Alias” particle handling p Interface to other generators (Pythia, PHOTOS) p n p Set of validation plots provided with the package n n p Already widely used for B-physics signal production Generator information for relevant particles Generator-level observables of most important decay modes Tests ongoing of use in inclusive event production 12
Remaining issues p Software-related: n CMS interface: p most issues solved, new implementation requires understanding: § CP violation p n Evt. Gen. LHC code: p p timing optimization? CMS users mostly complaining about lack of documentation: Physics-related: n Particle properties (including not just PDT standard information, like masses and lifetimes, but also B-specific, e. g. helicity amplitudes in P VV decays): p p n How often are they updated (LHC-wide versioning)? Compatibility with other MC generators? Tuning issues: p Decision of using/not using Evt. Gen as a general decay tool also depends on experience by other experiments: § Status of comparison with other generators § Effort needed to adjust MC tunes after use of Evt. Gen 13
- Slides: 13