1 / 9

Test67 for the verification of EM models

Test67 for the verification of EM models. Luciano Pandola INFN – LNGS. Intercomparison exercise. In 2007 a “comparison exercise” has been performed within the g -ray spectrometry community

stacia
Download Presentation

Test67 for the verification of EM models

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Test67 for the verification of EM models Luciano Pandola INFN – LNGS

  2. Intercomparison exercise • In 2007 a “comparison exercise” has been performed within the g-ray spectrometry community • Goal: compare detection efficiencies predicted by different MC codes for a given (simplified) geometry of interest • each participant submitted the results calculated with his/her favourite MC code • participants used Geant4, Geant3, PENELOPE, MCNP, EGS et al. • Appl. Rad. Isotopes 66 (2008) 764 • found differences up to 10% at very low energy. Typically < few % • Use this setup as a benchmark for the existing and new EM models in Geant4  test67

  3. Check EM physics with test67 • The same test67 job issistematicallyrunateachnew referencetag, usingseveral EM builders • 6 energiesbetween 45 keV and 2000 keV • 5·106 events/run • Convenient wayto: • test for anomalies/bugs, possibly introduced in the development • Cross-check the stability of results and the consistency of the EM builders • Benchmark the CPU performances of each builder and track for improvements • Results available at • /afs/cern.ch/sw/geant4/user/vnivanch/verification/electromagnetic/test67

  4. Stability of results vs. tag, G4EmPenelopePhysics Full peak efficiency at 120 keV (nominal geometry configuration) stable vs. Geant4-ref tag 09-04-ref-01 Penelope v2008 Full peak efficiency at 120 keV (%) 09-05-ref-09 Penelope v2001

  5. Stability of results vs. tag, G4EmStandardPhysics_option3 Full peak efficiency at 120 keV (%) 09-05-ref-09 09-04-ref-05 Major variation in result seen in 09-05-ref-09 feedback to Vladimir. Due to a bug, immediately fixed intags emstand-V09-05-53 and emutils-V09-05-43

  6. Benchmark of CPU performances At each new reference tag, the same test67 job is executed on pcgeant08@CERN using different EM builders. Absolute numbers are meaningless, but it is interesting to check for relative differences between different models and/or different Geant4 tags Clear improvement for Opt3 and Livermore between ref-06 and ref-07 (new Rayleigh scattering model)

  7. Backup

  8. Comparison with other codes • Results from Geant4-sim can be compared also with reference values obtained with other codes in the intercomparison exercise • error bars in the reference values are the spread of results provided by participants who used the same code • Good agreement with EGS, Geant4 (of course!), Penelope (of course!) Total efficiency at 120 keV Peak efficiency at 120 keV

  9. Summary • Test (test67) available in Geant4 to implement the intercomparison exercise of Appl. Rad. Iso. 66 (2008) 764 • well-defined geometry, typical use-case of g-ray spectroscopy • Results obtained with the dev version of Geant4 can be compared with the reference results reported in the intercomparison and obtained with different MC tools • way to quickly check for (big) bugs during the development • scripts and processing tools already available • Convenient way also to monitor the stability vs. reference tag of several parameters • not only efficiencies, but also fluorescence yield, spectra, etc. • The application is able to handle all sets of EM models, though more focus was given to Penelope, up to now • good performance of EM models (results and stability)

More Related