1. IEEE NSS 2012 IEEE NSS 2007 Honolulu, HI Best Student Paper (A. Lechner) IEEE TNS April 2009 Same geometry, primary generator and energy deposition scoring reused for the results presented in this talk Verification of consistency: published/new results in same configuration Geant4 8.1p02 Geant4 9.1

2. IEEE NSS 2012 High precision measurements intended for simulation validation Traditionally considered a reference for Monte Carlo codes

3. IEEE NSS 2012 Experimental set-up Beam energy: 25 keV – 1 MeV Incidence angles: 0 o, 30 o, 60 o Be, C, Al, Fe, Cu, Mo, Ta, U Be, C, Al, Ti, Mo, Ta, U Targets: [Sandia79] [Sandia80] Experimental uncertainties: 1.2 - 2.2% (nominal) Energy deposition profile Total deposited energy Measure:

4. IEEE NSS 2012 Simulation configuration infinite layer e - beam calorimeter front foil Sandia79: calorimeter placed at different depths Calorimeter and front/infinite layers: same material Geometry : as in experiment Mass geometry + readout geometry Geometry : as in experiment Mass geometry + readout geometry e/  physics low energy EEDL/EPDL (“Livermore”) low energy – Penelope Standard e/  physics low energy EEDL/EPDL (“Livermore”) low energy – Penelope Standard Multiple scattering Urban* Goudsmit-Saunderson Multiple scattering Urban* Goudsmit-Saunderson Step limitation 1, 10, 1000  m no step limitation Step limitation 1, 10, 1000  m no step limitation Secondary production threshold 250 eV (low energy) 1 keV (standard) Secondary production threshold 250 eV (low energy) 1 keV (standard) Sandia80: whole volume is sensitive Geant4 versions 8.1, 9.1, 9.2, 9.3, 9.4, 9.5, 9.6-  (with latest patch applied) Geant4 versions 8.1, 9.1, 9.2, 9.3, 9.4, 9.5, 9.6-  (with latest patch applied)

5. IEEE NSS 2012 Validation analysis Compatibility of experiment-simulation:  2 test  2 test case = target material, beam energy, beam angle Significance of the  2 test:  = 0.01  G4model = N  >0.01 N tot efficiency of a Geant4 physics model fraction of test cases in which simulation is compatible with experiment Categorical analysis to determine whether two Geant4 physics models, or two Geant4 versions differ significantly in accuracy Categorical analysis to determine whether two Geant4 physics models, or two Geant4 versions differ significantly in accuracy Contingency tables whole sample matched pairs Fisher’s exact test Barnard’s exact test Pearson’s  2 test  2 test with Yates continuity correction McNemar’s test (matched pairs)

6. IEEE NSS 2012 Energy deposition profile C, 1 MeV Geant4 9.6-  similar to 9.5 Most accurate Geant4 version: 9.1 Most accurate Geant4 model: Livermore-9.1 Most accurate Geant4 version: 9.1 Most accurate Geant4 model: Livermore-9.1 p-value Geant4 9.1-9.5 equivalence: 0.008 p-value Livermore-Penelope: 0.002 p-value Livermore-Standard: <0.001

7. IEEE NSS 2012 Total energy deposition Efficiency vs. Geant4 version in a single volume of elemental material (the simplest test case one can think of) Dependency on step limitation observed in Geant4 9.4

8. IEEE NSS 2012 Multiple scattering models G4 Version Range Factor Step Limitation Lateral Displacementskin geom Factor ModelProcess 8.1. p02 0.023.5G4UrbanMscModelG4MultipleScattering 9.1. p03 0.02fUseSafetyTRUE02.5G4UrbanMscModelG4MultipleScattering 9.2. p04 0.02fUseSafetyTRUE32.5G4UrbanMscModelG4MultipleScattering 9.3. p02 0.04fUseSafetyTRUE32.5G4UrbanMscModel92G4MultipleScattering 9.4. p04 0.04fUseSafetyTRUE12.5G4UrbanMscModel93G4eMultipleScattering 9.5. p01 0.04fUseSafetyTRUE12.5G4UrbanMscModel95G4eMultipleScattering 9.6. b01 0.04fUseSafetyTRUE12.5G4UrbanMscModel95G4eMultipleScattering Further tests in progress to evaluate the effects of different parameter values and models in the Sandia79/80 test configuration

9. IEEE NSS 2012 O. Kadri, V. Ivanchenko, F. Gharbi, A. Trabelsi Incorporation of the Goudsmit–Saunderson electron transport theory in the Geant4 Monte Carlo code NIM B, Vol. 267, no.23–24, pp. 3624–3632, Dec. 2009  2 test Geant4 9.3p02 (Sep. 2010): p-value = 1.6 10 -34 Geant4 9.5p01 (Mar. 2012): p-value = 1.8 10 -22 Experimental data from Sandia-79 report Goudsmit-Saunderson multiple scattering model Al 521 keV G4GoudsmitSaundersonModel - for electrons and positrons 9.3p029.4p049.5p01 G-S0.13±0.060.17±0.06 Urban0.27± 0.080.23±0.080.47±0.09 Efficiency (with “Livermore” e/  models)

10. IEEE NSS 2012 Penelope 2001-2008 re-implementation Geant4 9.1: Penelope 2001 Geant4 9.5: Penelope 2008 (default) Penelope 2001 (available) Model Geant4 9.1 p03 Geant4 9.5 p01 Penelope 20010.30 ± 0.080.20 ± 0.07 Penelope 2008-0.13 ± 0.06 EEDL-EPDL0.73 ±0.080.47 ±0.09 Penelope 2008 model does not appear to have improved Geant4 simulation accuracy w.r.t. Penelope 2001, nor w.r.t. EEDL/EPDL (“Livermore”) models