Created attachment 394 [details] Simulation file, PhysicsLists and excel-file comparing the output I have tried to simulate the transmitted signal of a transmission scanning electron microscope for incident electrons of 30 keV with different Geant4 versions (10.0p04, 10.1p03, 10.2 and 10.2p01). My problem here is that the signal differs so dramatically when using different MSc processes and also different Geant4 versions. In the experiment I simulate, 30 keV electrons are only transmitted through a carbon substrate of 12.5 nm thickness and electrons scattered up to an angle of about 19 mrad are detected and contribute to the signal. In other words, electrons scattered more than 19 mrad are not detected. I have 2 major concerns here - please have a look at the attached Excel-file, where the highlighted numbers are reasonable results: 1) In V10.0p04, the use of the G4UrbanMScModel plus provided a good signal value of 0.86, but in the new versions 10.2 this signal is reduced to 0.22, which does not make sense for this geometry! 2) I also tried to switch off the G4eCoulombScatteringModel and hence the single scattering contribution. There was an effect in V10.0p04 but in versions 10.1p03 and above this model has absolutely no effect on our signal, which seems not right to me. We appreciate your help in this matter as we hope to continue using Geant4 for these kind of simulations! Please let me know when you need further information on this problem. With best wishes, Marion
Thank you for reporting. Your issue is now assigned to the coordinator of the Standard Electromagnetic physics working group.
Hello, sorry for the late start looking into the bug report. What I see for 10.2 the Physics List is instantiated at a wrong moment - at the end of the PhysicsList.cc you have unnecessary lines ConstructParticle(); ConstructProcess(); this should be removed. ALso the default name of a EM builder in class constructor should correspond to the default constructor. Please, fix above problems and try to get new results. VI
Created attachment 407 [details] new PhysicsLists and Results
(In reply to comment #2) > Hello, > > sorry for the late start looking into the bug report. What I see for 10.2 the > Physics List is instantiated at a wrong moment - at the end of the > PhysicsList.cc you have unnecessary lines > > ConstructParticle(); > ConstructProcess(); > > this should be removed. ALso the default name of a EM builder in class > constructor should correspond to the default constructor. > > Please, fix above problems and try to get new results. > > VI Dear Vladimir, thanks for your reply. I have changed the PhysicsList according to your comment (see attachment). Results have changed to about 0.65 for G4EmStandardGS and G4EmStandardSS (see column G in excel file). This value is different to any results before but would be ok for a comparison to the experiment (here we measure a signal of 0.7 - 0.75). Using the Urban MSc model via G4EmLivermorePhysics, G4EmStandardPhysics_option4 or G4EmPenelopePhysics leads again to 0.22, which is far too low. Marion
Hello Marion, with the recent Geant4 version 10.3 I can confirm that for your setup and condition to transmitted electrons we have: Urban msc (Op3, Op4) - 30% WVI - 86.9% GS - 88.9% SS - 88.2% WVI, GS, SS for such thin target are working in a similar way when they switched to single scattering mode, so have very similar distributions and are recommended to be used for such use-case. Urban model has instead small step approximation of multiple scattering which fails for very small steps. So, we would propose to use SS physics list or GS physics list for your case. Their results should be stable between Geant4 releases. The fact, that the Urban model in past provided different result only means that it is unstable but never assumed to be correct in this use-case. VI
The same test for 10.2p03: Urban msc (Op3, Op4) - 30.9% WVI - 85.6% GS - 89.0% SS - 89.3% So, we have stable results and GS and SS can be recommended for this use case. VI
