Inconsistent results for multilayer shielding

I am simulating multilayer radiation shielding.

To test my geometries I set all layers to the same material, but I get inconsistent results.
I just ran a test in which I simulate stacks of aluminium plates with the same total thickness but different numbers of layers and plot the total dose detected behind the shielding versus the number of plates.

The result is roughly linearly increasing total dose results even though the total thickness of the shielding stays constant.

The plot below shows my results for simulating an aluminium shield that is split into two layers. The total thickness stays constant, but the ratio between the two layers changes over the range of the x-axis.
I would expect to see a flat line with no trend, but why am I getting a sinusoid pattern?

Could it be that there are some boundary layer effects?
Or does Geant4 have issues when a layer thickness is on the same as the free path length of the particle?

to try to reproduce the problem, we need more informations :
nature and kinetic energy of incident particles
transverse size of the layers
transverse size of the beam

Gradient-Test.txt (3.2 KB)
Electrons500keV.txt (1.4 KB)
beamMacro.txt (281 Bytes)

Thank you Maire for your reply!

I attached my GDML geometry file, my particle macro and the beam macro.
Is there anything else you need to replicate the simulation?

The transverse size in that specific simulation was 10m, but for planar geometries, this is irrelevant as long as the transverse size is several orders of magnitude higher than the thickness.

I repeated your exercise with example TestEm3, at 2 energies (1 MeV and 5 MeV),
and for { 1, 10, 50, 100 } layers. I see effects of ~2 - 4 % on energy leakage.
To improve stability, we recommend EmStandard_option4 : it has been designed for that purpose ! But it is slower.
If you have time, please redo your exercise with option4 and post the result. It can be of general interest.

anton.mac.txt (432 Bytes)

1 Like

Thank you very much!

So you are saying that this is a known bug of the simulation software?

Sadly I am already in the final weeks of my master’s thesis and I am using a Geant4 application written by my supervisor.
I will tell him about this, but I currently cannot test your proposed solution.

Is there anything I could cite for my thesis about this known bug?
Otherwise, I just have to write that I observed this weird effect and EmStandard_option4 is reccomended by a member of the forum.
I don’t really know how to write that in a thesis document.

It is not a Geant4 bug.
In transport Monte Carlo codes, there are basic problems in algorithms of so called condensed history processes (eg. multiple scattering, stopping power), in opposition to analog processes (eg. single Coulomb scattering …).

To see the origin of the problem, and how it has been solved in EGSnrc, start with I. Kawrakow, Med. Physics. 27-3 (2000) 499

To know how it has been implemented in Geant4, ask precise references directly to

There is a program test : examples/extended/medical/fanoCavity

Ok, I did not know about this.
The publication I. Kawrakow (2000) is very interesting.
Now I have something that I can talk about in my thesis.
Thank you very much.

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.