I have stumbled on this problem when trying to do a consistency check.
The test consists of shooting a infinitely small proton beam onto a thin G4_WATER slab (0.01mm thickness and very large lateral size). The world is made of G4_Galactic. I am using via a modular physics list only G4EmStandardPhysics_option4, G4DecayPhysics, G4RadioactiveDecayPhysics, G4EmExtraPhysics, G4HadronElasticPhysicsHP, G4StoppingPhysics and G4IonPhysics. So no inelastic nuclear interactions. I have also set the production cut of secondaries to 1*km to avoid creating any secondaries. I have verified that I have no secondaries and that the protons do exactly one step in the slab, at the boundaries of the volume (no user step limit). So nothing escapes.
I score the total energy deposit in the slab for a number of primaries via GetTotalEnergyDeposit from each step. I also score the energy spectrum of the protons in the slab with 10keV energy bins, every entry I add to the histogram is weighted by the length of the step. Then I get from the EMcalculator the unrestricted (1*km cut) dEdx for protons in G4_WATER with 10keV energy steps. Finally I try to calculate the energy deposit in the slab by a bin-wise multiplication of the spectrum with the dEdx corresponding to that proton energy bin. The result I get is that the energy deposit calculated from the spectrum is always about 1.5% lower than what I get from the GetTotalEnergyDeposit. Even when the energy spectrum is scored using the post step energy (which would give the highest dEdx) then still the same discrepancy occurs. I would need to shift the spectrum by a whole 4MeV to match the two energy deposit calculations.
I understand that for the energy loss along a step, fluctuations are added to the energy loss, but I thought that for large statistics the result should converge to the mean value calculated from the EMcalculator with an accuracy better than 1.5%. Is there an easy way to switch of these fluctuations? Does multiple scattering play a role in this test?
Thanks for any help or hint