The geometry is a thin layer of homogenous material ,
the incident particle is an electron.
The setup is shown on the picture
with 10 superimposed events: 530 micron of silicium, e- 1 MeV.
The most probable energy loss of an electron of energy 1 MeV in the Si layer
is around 200 keV. However, due to the multiple scattering and delta ray
production the primary electron can deposit more energy or even it can be
completly absorbed in the detector (in about 4 % of the cases).
The distribution of the energy deposit is a very sensitive test of the
simulation of the multiple scattering and of the energy loss computation
(dE/dx and its fluctuation).
GEANT4/GEANT3/data comparison is shown here,
where 1 000 000 events have been simulated, the production threshold
is 7 micrometer (the corresponding energy for e-
in silicon is 10 keV).
The experimental data have been taken from the paper of M.J.Berger et al.,
Nucl.Inst.Meth. 69 (1969) 181.
Remark 1. The simulation results of GEANT4 (and GEANT3) are very stable, they change very little with the production threshold. (see edep spectrum with prod thresholds of 1 mm and 7 micrometer) .The small difference at low energy comes from the backscattered electrons.In the case of big production threshold the electron goes through the silicon in just one step, there is no backscattering practically. Using the smaller production threshold , the step is limited by the delta ray production , there are more steps in the silicon and the probability of the backscattering is bigger.
Remark 2. The peak around 1 MeV (absorption peak) seems to be too high in the simulation because the simulation did not take into account the experimental errors. If the noise of the detector is simulated with the parameters given in the experimental paper, the absorption peak is quite close to the data. (The noise is just one component of the experimental error, of course.)
GEANT4/GEANT3/data comparisons are shown here for target thicknesses
0.10 g/cm2 (370.37 micrometer)
and 0.22 g/cm2 (814.82 micrometer)
The number of simulated events is 100 000, the production threshold is
6 micrometer (10 keV in energy for electrons).
The experimental data have been taken from D.H.Rester and J.H.Derrickson
Journal of Appl.Phys. 12(1970)714.
The difference between the simulated and experimental spectra comes mainly from the experimental errors (see the spectrum with simulated errors on the plot 0.10 g/cm2 + errors, where the estimated error was used in the simulations).
For the case of 0.10 g/cm2 target thickness both Monte Carlo reproduce the data fairly well, while in the case of 0.22 g/cm2 thickness GEANT4 is closer to the data, GEANT3 gives too small transmission (or too big energy deposit in the target).
The code of these examples are in geant4/examples/extended/electromagnetic/TestEm5