1. GRAS Validation and GEANT4 Electromagnetic Physics Parameters R. Lindberg, G. Santin; ronnie.lindberg@esa.int Space Environment and Effects Section, ESTEC

2. 2 Presentation Outline Introduction A few Words About GRAS and MULASSIS GRAS Internal Validation Comparison with MULASSIS GEANT4 Electromagnetic Physics Tuning the parameters in GRAS GRAS applied to complex geometry: ConeXpress Conclusions

3. 3Introduction ConeXpress radiation analysis ESABASE Ray-tracing and SHIELDOSE-2 curve GEANT4 Ray-tracing (SSAT) and SHIELDOSE-2 curve Used following tools for comparison GRAS Developed by G. Santin and V. Ivantchenko Uses GDML geometry; modular physics Modular analysis driven via script SSAT Developed by Qinetiq Ray-tracing (a.k.a sector shielding analysis) MULASSIS Developed by Qinetiq 1D multi-layer geometry.

4. 4 ConeXpress Results GEANT4 SSAT ray-tracing results agree with ESABASE However, GEANT4 GRAS full Monte Carlo gives very different results (orders of magnitude) Uses same geometry model as SSAT analysis First validation attempt GEANT4 internal comparison GRAS ↔ MULASSIS Shows discrepancy of ~20 % for a semi-infinite slab case Greatest difference in lower energy range (≤ 2 MeV) for electrons

5. 5 Understanding the Problem (1/3) The geometry setup used was the semi- infinite slab case 2 mm Silicon target 3 mm Aluminium shield Dose in energies below 1.5 MeV comes from gamma radiation e - -contribution starts to dominate around 1.5 MeV 3

6. 6 Understanding the Problem (2/3) GRAS analysis was inserted into MULASSIS to obtain e- and gamma cont. Gamma contribution agrees well between the two. Simulations show that there were differences in the e- contributions between GRAS and MULASSIS

7. 7 Understanding the Problem (3/3) Dose from gamma-contribution is the same but... …e- contribution differs and… …statistical errors are small (<1%) compared to total dose value, so difference is not due to statistical error, furthermore… …the difference in dose between GRAS and MULASSIS is largest at “threshold energy”, so… …what’s the catch?

8. 8 Electron EM Processes and Fine Tuning Same EM physics used in GRAS and MULASSIS Cause of different results was due to “fine tuning” of the electromagnetic energy loss modelling Several parameters influence the modelling of GEANT 4 EM: facRange: Maximum fraction of kinetic energy that particle can loose in a step Integral: If true, dE(step) is obtained with integral of dE/dx curve Cuts: Is the production cuts for secondary electrons StepMax: Is one of the most important. Limits the maximum step length. “Process” in GRAS. This parameter is not available in MULASSIS

9. 9 Internal Validation Conclusion (1/2) GRAS gives near perfect agreement with MULASSIS when using the same EM physics parameter Integral set to true facRange set to 1.0 stepMax set to 100 mm (similar to not having stepMax at all)

10. 10 Internal Validation Conclusion (2/2) Several runs were conducted to verify correlation E.g. Sphere case, maxtheta=90, protons and electrons, Notice the scale.

11. 11 EM Physics Tuning – Parametric Study Parametric study to look at effects of different settings Parameter ranges: facRange: 0.2 - 1. Integral: Boolean – true or false Cuts: between 0.01  100 mm StepMax: between 0.01  100 mm (100 mm ~ no step limiting)

12. 12 Parameter Comparison (1/2) Dose differs 2.5x depending on StepMax

13. 13 Parameter Comparison (2/2)

14. 14 Tuning Effect with Space Env. Spectra Ran simulations in GRAS for different spectra and Al shielding thickness: e- GTO e- MEO (Galileo) e- GEO p+ GEO MULASSIS simulated by using StepMax=100.00 mm and StepFunction=1.0

15. 15 Tuning Effect with Space Env. Spectra Trapped e- GEO spectrum Average dose per event (MeV) Al. thick. mmGRAS e- MULASSI S e- GRAS/MUL 30,029830,02410124% 40,008180,00653125% 50,002810,00228123% 100,000380,0004195% Trapped e- MEO spectrum Average dose per event (MeV) Al. Thick. mmGRASMULASSIS GRAS/MUL 30,048240,04066119% 40,014580,01167125% 50,004740,00363131% 100,000450,0004697% Trapped e- GTO spectrum Average dose per event (MeV) Al. thick., mmGRASMULASSIS GRAS/MUL 30,053910,04625117% 40,017360,01399124% 50,005910,00456130% 100,000470,0004996% solar proton GEO spectrum Average dose per event (MeV) Al. thick. mm GRAS e- MULASSI S e- GRAS/MUL 31,88 99,8% 45,585,57100,1% 54,124,1399,9% 103,333,3499,8%

16. 16 Next Step – Complex Geometry Currently conducting analysis on complex geometry – ConeXpress Use radiation spectra from SPENVIS Run each particle spectra separate and combine to obtain total ionised dose. Presents different problems than simple geometry Number of simulated events has to be very high due to thick shielding generated by subsystems, especially for electrons

17. 17 Next Step – Complex Geometry GDML model of ConeXpress

18. 18Conclusions Internal validation (GRAS ↔ MULASSIS) successful Earlier difference due to different physics parameters GRAS Parametric study of EM physics parameters shows difference Up to 30%, using a space environment spectra Up to 2.5 times, using mono-energetic beam particle source Tentative set of parameters chosen as facRange to 0.2 Integral set to true Cuts around 0.01 mm StepMax around 0.1 mm – trade-off between CPU time and small step size  impacts radiation analyses results Suggested implementation of StepMax and facRange in MULASSIS

19. 19 Integral Parameter Integral integrates dE/dx to get smoother curve Rate of energy loss, dE/dx, changes with energy Integrated area is total energy lost in step dx dE/dx 1 step

20. 20 facRange and Integral Parameter dE Energy facRange limits dE to fraction of original energy Low fraction lost High energy lost All energy lost