When using a geometry constructed with the G4PVReplica method, the Transportation process sometimes returns a 0 step length and the particle remains stuck at the same place forever. I wrote a small example to reproduce the problem in ftp://guest:time@ftp.metas.ch/voros/totabs.tgz It is a water cylinder surrounded by air, and the cylinder is divided into Z-Phi-R replica. For each event an e- of 5.332 MeV located at -3.2 cm is shot along the Z axis towards the cylinder surface located at 0 cm. I tried to avoid "usual" errors with replica (i.e., not filling the entire mother volume, etc.), so this must be something else, and DAVID/DAWN did not detect any overlapping volume. I'm using Geant4.6.0.p01 on a Linux RedHat 9.0 PC. This small example works similarly as the Geant4 examples : define the usual Geant4 environment variables, then 'cd' to the 'totabs' directory, 'gmake' and execute it. The vis.mac file will be executed first (it uses the OGLIX viewer), and then you can give the command '/run/beamOn 8'. I get the "desired" problem in the last event: Step# X Y Z KineE dEStep StepLeng TrakLeng Volume Process 0 0 fm 0 fm -3.2 cm 5.33 MeV 0 eV 0 fm 0 fm World initStep 1 0 fm 0 fm 0 fm 5.33 MeV 6.91 keV 3.2 cm 3.2 cm World Transportation 2 0 fm 0 fm 0 fm 5.33 MeV 0 eV 0 fm 3.2 cm AbsorberRPhiZDiv Transportation 3 0 fm 0 fm 0 fm 5.33 MeV 0 eV 0 fm 3.2 cm World Transportation 4 0 fm 0 fm 0 fm 5.33 MeV 0 eV 0 fm 3.2 cm AbsorberRPhiZDiv Transportation 5 0 fm 0 fm 0 fm 5.33 MeV 0 eV 0 fm 3.2 cm World Transportation ........ So the particle seems to oscillate between the World and AbsorberRPhiZDiv (the replicated) volumes in this case, but I also saw it remain stuck at an other boundary between two replica when using a slightly different setup. I also tested a couple of changes in the geometry, like R-Z-Phi replica instead of Z-Phi-R, using G4PVDivision instead of G4PVReplica, changing the granularity in the totabs/src/totabsDetectorConstruction.cc file, but the problem remains the same. Thanks a lot for any help. SĂĄndor
Dear Sandor, We have been able to reproduce the problem, and have created a protection to enable the Navigator to recognise problems of this type and recover/continue navigation after a number of zero steps. You will find this in Geant4 6.2, which was released last Friday. We plan to continue to investigate this, in order to discover the source of the problem, in order to correct it. Thank you for your report, and especially for the very useful test case. Best regards, John Apostolakis.
The problem is under investigation. It has been observed that the loop (zero-step length with particle bouncing between two volumes) is reproduced only in the case when the combination replica-in-Phi + replica-in-Rho is used (i.e. tubs sliced in Phi, and each phi-slice is sliced in Rho).
The problem has been fixed in the development tag "geomnav-V06-02-00". The fix will be included in the next patch or release. Thanks for having reported the problem.
