For secondary electron emission modelling using GEANT version 10.05 patch 1 and either the default, Penelope, or Livermore models, I observe that if incident electrons are at large angles from the surface normal (e.g. 80 degrees from normal, or equivalently 10 degrees from grazing), then the incident electron is simply elastically scattered from the surface. The net effect is a secondary electron yield of unity at these angles, which is physically incorrect. In case these details are important: I am looking at incident energy ranges of 10 keV to 30 MeV, and the pre-defined GEANT4 materials of Copper, Stainless Steel, Aluminum, Aluminum Oxide (Alumina), Niobium, Beryllium Oxide, and Graphite. The QGSP_BERT hadronic processes list was used, but I do not believe that these processes are relevant at the incident energy ranges considered.
Hello Sebastien, please, assign the problem to me (later I may re-attach it to somebody else). Cheers, Vladimir
Thanks Vladimir, reassigned.
Hello, Scott, I agree with you, that likely the problem do not depend on hadronic part of Physics List and on concrete material. At this energies only Urban multiple scattering model is responsible for backscattering. The alternative may be single scattering of Goudsmit-Sounderson models - SS and Opt4 Physics Lists. Both should be more accurate but significantly slower than the default EM physics. Can you try, at least, Opt4 (FTFP_BERT_EMZ)? Is it difficult for you to switch to Geant4 10.7? VI
Hi Vladimir, The project has ended for which we were using GEANT to generate data, and I do not have funding to request the appropriate person on my team to look into this further. I mainly wanted to alert the GEANT developers to a possible bug. If it may help, I could send you some GEANT input files which produced the results we observed. SR
Hi again Vladimir, My group was able to collect more data on my end: when we used Opt4 (FTFP_BERT_EMZ) withr both version 10.5 and 10.7, we get the same behavior as what was originally reported for this bug: for an incident angle of 80 degrees from normal, the secondary electron yield curve is unity for all incident energy values above the model's energy threshold.
Created attachment 663 [details] electron backscattering coef.
Comment on attachment 663 [details] electron backscattering coef. Hi Scott, We have investigated the reported problem by: - taking a fresh Geant4-10.7 - modify a bit `geant4.10.07/examples/extended/electromagnetic/TestEm5` especially the `PrimaryGeneratorAction` in order to be able to change the (default) normal incidence angel of the primary particles We obtained the attached results so we could not reproduce the reported problem.
Hi Mihaly and Vladimir, The attached plot shows the backscattering yield, which is defined to be only emitted electrons with an energy greater than 50 eV. (See the "Experimental" section of the Neubert and Rogashchewski paper cited in the plot, and note their 50 V bias, and also the previous literature they cited is more explicit about this.) This is different from the total secondary electron yield, which is the yield of electrons of all energies. (The literature unfortunately does not always use consistent terminology, and sometimes "secondary electrons" mean only electrons with less than 50 eV of energy, and other times "secondary electrons" means electrons of any energy.) When I run GEANT4, I observe a total secondary electron yield of unity for an incidence angle of 80 degrees from normal. Are you able to observe the total yield (not just backscattering) with your simulation set-up? I will try to duplicate what you have produced based on your description. If there are any files you may want me to upload from my simulation set-up, please let me know and I can try to post them here.
I just realized something after posting: due to the energy production thresholds in GEANT, GEANT will not be producing electrons less than 100 eV (if I remember the Livermore production energy correctly). So the backscattered electron results in the literature is a valid comparison. I still do not understand why I get a yield of unity, while you are showing something different. I will look more into this on my end and report back when I know more.
Created attachment 664 [details] updated 60 [keV] e- bsc Hi Scott, We cannot expect an accurate secondary electron yield simulation with Geant4 using any of the standard, (neither the penelope nor livermore) EM constructors. Without going too much into the details, we need to keep in mind the (several) limitations posed by the condensed history simulation technique utilised when these physics lists are used. These (and the models utilised) will prevent to perform accurate simulations of low energy (E < ~2-3keV) electron transport (while the majority of the secondary electrons escape with a very low < 20-30 [eV] energy). Note, that the goal was to simulate the "backscattering electron yield"! According to your report, all electrons elastically backscatters form the surface when the angle of incidence is higher than 80 [deg.]. That would correspond to a backscattering coefficient of 1.0 (at least if no any secondaries escape the surface). As we can see from the plot, this is clearly not the case. In fact, the plot was generated by setting the lowest electron energy (a kind of e-/e+ tracking cut) to 50 [eV], the low energy limit of the secondary electron production threshold to 50 [eV] with a very low (1.0E-3 [um]) secondary e- production threshold (to ensure that this 50 [eV] low energy limit is reached). /cuts/setLowEdge 50.0 eV /process/em/lowestElectronEnergy 50.0 eV /testem/phys/addPhysics emstandard_opt4 /run/setCut 0.001 um /run/initialize These might bring the simulation conditions closer to that of the experiment. (Note however, that these do not guarantee anything regarding correctness.) Moreover, one can extend TestEm5 such that the correct e- backscattering yield is collected (collect all e- that escaped the surface in the backward direction). For completeness, you can find an updated plot attached generated by TestEm5 with all these modifications and settings mentioned above, using the penelope EM physics constructor /testem/phys/addPhysics empenelope instead the opt4. It is because its model for e- ionisation is more suitable for such low energies than the livermore one used in opt4 while the most critical part, the model for e-/e+ Coulomb scattering is the same (GS-MSC model). Cheers, Mihaly
Hi Mihaly and Vladimir, I have not had an opportunity to dig deeper into this, and I do not anticipate having more time to do so in the foreseeable future. If you are satisfied that GEANT is behaving as expected, we can close this ticket. Thanks for looking into this.
