Problem 1027

Summary: fixed structure in electron field when shifting whole geometry and source
Product: Geant4 Reporter: Bjoern Kreisler <bjoern.kreisler>
Component: processes/electromagneticAssignee: Vladimir.Ivantchenko
Status: RESOLVED FIXED    
Severity: normal CC: asai
Priority: P3    
Version: 9.1   
Hardware: PC   
OS: Linux   
Attachments: source files and example image of the problem in tar archive

Description Bjoern Kreisler 2008-10-06 11:12:49 CEST
Created attachment 31 [details]
source files and example image of the problem in tar archive

1) detailed description
there is a peak in the x-profile which is located at x=-epsilon even when the geometry and the source are moved away from the centre.

2) setup
world is filled with G4Air.
two boxes(Aluminium or Tungsten) are part of the world.
electrons are emitted from a source (circle with gaussian divergence)measurement volume (also air) is between source and the two boxes.

3) result
in the x-profile in the electron field appears a peak at x=-epsilon, although the source is evenly distributed. the peak is still at x=-epsilon when the whole geometry (source and boxes) is shifted. the peak is not just a single wrong histogram entry, but it has a width.

4) additional information
the problem was discovered with the geant4.9.1 but is still present in the new installation with the patch03.
even moving the two boxes further away from the measurement volume leaves the peak unchanged.
the problem is not in direct connection to the G4EMLOW, as I also tried the standard physics.

5) attachment 
there is a tar archive (081006_kreisler.tar.gz) which includes the following:
 xprofile.eps - plot of the x profile for two different setups
 vis.mac
 GNUmakefile
 testsim_root.cc - main source
 src/
 include/

the compiled program needs the following start script:
testsim_root 7 500000 filename 1
[7= energy (in MeV), 500000 = number of electrons from source, 1 = number of the simulation]
Comment 1 Gabriele Cosmo 2008-10-07 02:16:32 CEST
User's application requires analysis, in the way by which:
a. primaries are generated;
b. quantities are stored from sensitive detector and analysed. 
Wrongly assigned to geometry; being assigned to EM.
Comment 2 Bjoern Kreisler 2008-10-29 14:04:58 CET
I tested the setup with the Penelope physics and even with the Geant4.9.2-b01, but the strange peak in the x-profile is still present.
Can I help with anything else?
Comment 3 Vladimir.Ivantchenko 2008-11-19 14:46:36 CET
Dear Bjoern,
Sorry for long study of the problem. It was connected with interaction between multiple scattering model and geometry. Your setup emphasize it in sucha way that it become seen. The complete solution of the problem needs more work.

The work around is simple: in your detector construction there is definition of the world logical volume. Add last argument as "false":

  experimentalHall_log = new G4LogicalVolume(	experimentalHall_box,
						Air,
						"expHall_log",
						0,0,0,false);
by this flag you disable voxelisation of the world volume. My estimates that this will not affect CPU performance of your applications. PLease, chech that the artifact disappears.

Vladimir