Energy Deposition Histogram Boundary Effects
Issue #90
on hold
The energy deposition histogram shows boundary effects under several particle types and physics lists. Around the boundary of similar material, greater energy is seen before the boundary and less afterwards.
Comments (9)
-
reporter
-
reporter
Step length has been added to the precision output to allow easier debugging. This lead to a fix for user limits step length in the collimator base class.
-
reporter
Long steps that could potentially appear to cross boundaries are fixed with user limits. Steps can't cross boundaries in Geant4 so it was long transverse steps. Still, this should be fixed in the collimators. This small issue fixed in commit #7808096.
-
Range filling has been implemented, and the effect is much reduced but not gone. Shown is the Energy loss from the 250 GeV electron copper collimator example:
-
Laurie suggested to check energy loss against z instead of s
-
reporter
4TeV Proton, em_standard physics into 1m closed collimator. Used precision output. Definitely some edge effect with s coordinate as can be seen from graphs - investigating.
-
reporter
Further investigations with:
proton, 4TeV, 1m Cu x 2 one after the other with no gap, closed collimators with different ranges showed no correlation with range. 1mm, 1m, 10m, 100m, 10km were tried and no effect was seen. The 'old' (non-modular physics lists QGSP_BERT was used.
There is a slight boundary effect with S and (global) Z coordinates. This was tested previously in detailed when this was developed and was as accurate as can be. With histogram bins not much smaller than the length of the object, these are not seen on average.
-
reporter
This is mostly solved. If you create histogram bins much narrower than the range cut lengths, you will inevitably start to see boundary effects by construction. The key is to have your range less than the histogram bin width.
Additionally, the energy deposition is now attributed to a randomly chosen distance along the step with uniform probability. With sufficient statistics, this should give more accurate results.
Furthermore, the new rootevent output format has histogram error bars correctly calculated (as opposed to the old root format).
-
reporter
- changed status to on hold
Some examples / tests need to be run to confirm the described understanding. On hold for now.
- Log in to comment
From the track information we use the pre and post step points. We typically assign the energy deposit to the post step point as this empirically gave the energy deposition histogram with the least artefacts. Obviously, some still remain depending on the length scale of the histogram bins, the ranges and energy of the particles.
Geant4 does not explain clearly where the energy deposition should be attributed (ie which point, or the mean). After discussing with Jochem, we will always experience boundary effects when the histogram bin width is smaller than the typically step lengths involved. Therefore, if a very long track is found, we should split it and evenly divide the energy according to the histogram bin widths.
Jochem will implement a new fill method in BDSHistogram to achieve this. Also, we identified a series of significant and easy optimisations for the histogram class that will be implemented.