HPGe simulation in close geometry

Hi, We recently bought an HPGe detector in our lab and we were trying to compare the experimental efficiency with the one provided by the manufacturer software (ISOCS) and the one simulated in Geant4.

So I first made the geometry of our detector thanks to the manufacturer plan and at the beginning I just stored the energy deposition (if the Edep by step was > 0.6eV) in my crystal (minus the dead layer).

I made it for a punctual source of 152Eu at 1 meter : I recovered the efficiency at each peak and compared it to the one given by ISOCS and Geant4

As you can see, the results were quite close (deviation < 1%).

As it worked for a punctual source, I tried with a surface source (Cs137) consisting of a square surface (10x10cm) where the source was uniformly placed and covered by a thin layer of aluminum (about 0.3mm).

I made 2 measurements: one at 4mm from the entrance and one at 40cm during 30 and 60 minutes and reproduced them in ISOCS and Geant4.
And this is where the problems start:

At 40cm I get a deviation of about 3.9% with Geant4 and 5.6% with ISOCS for the efficiency (compared to the experimental result).

But at 4mm it gets worse: I have a deviation of 42% with Geant4 and 16.5% with ISOCS.
I tried to add the integration time in the simulation (Edep is stored if the time of the step minus the time of the first energy deposition is less than the integration time (about 1µs)) but same result.

/gps/pos/type Surface
/gps/pos/shape Square
#detector surface at 22.1mm 
/gps/pos/centre 0. 0. 26.1 mm
/gps/pos/halfx 5. cm
/gps/pos/halfy 5. cm
/gps/particle gamma  
/gps/ang/type iso
/gps/ene/type Mono
/gps/energy 661.7 keV

If anyone have an idea about this I will be glad to hear about it !

Thank you,


_Geant4 Version:_11-00-patch-03
_Operating System:_Ubuntu 20.04.6

Hello, I think it is a good agreement considering that the detector is ISOCS-characterised, which you show in the graph. It is a minor but whether any GEB (Gaussian Energy Broadening) is applied to the results.

Also, I see that you used the GPS macro for the primary generator. It is fine as it is but for tailored results, I have been making my own primary generation class and particle gun. However, if it is a calibrated gamma source like Eu152, I think it is okay. :slight_smile:

My last thought is also whether you have plotted a graph with uncertainty. With that, you can see whether the simulation results are within the experimental uncertainty which I think it would.

Good luck

Dear @ThibFe ,

this is more food for thought than an answer… but have you checked your experimental spectra at the different distances? Nevermind the simulations for now… see if the activity of the source starts factoring in and you get slightly not Gaussian peaks, pile-ups, etc… Geant4 on its own would never produce pile-ups and let’s not even start with the smearing of the peaks since that is entirely the users’ responsibility.