Message: Re: Step crosses a geometrical boundary and does not stop (from Run and Event forum) Not Logged In (login)
 Next-in-Thread Next-in-Thread
 Next-in-Forum Next-in-Forum

Question Re: Step crosses a geometrical boundary and does not stop (from Run and Event forum) 

Forum: Geometry
Re: Question Step crosses a geometrical boundary and does not stop (from Run and Event forum) (Ioannis Sechopoulos)
Date: 28 Apr, 2010
From: John Apostolakis <John Apostolakis>

Thank you for your clear message. What you report is unexpected, and I have no simple explanation. You may have uncovered a problem, likely in a solid, but I would like to double check some things you report - and suggest some additional checks to make. There is always the possibility of a badly formed geometry (at the lowest level or some other level above that) which could lead to different results.

0) Could you please provide the parameters of the key solids (either posting to hNews or in a direct email) so that we can double-check the relevant solid(s) as you defined them.

You wrote: > "I know that physiBreast finishes at Z=61 mm and that Z=62.4 mm is in "physiSkin", so the latter output is the correct one."

This is very suspicious - if your understanding of the geometry description is correct. It would seem that an intersection is missed by the solid's DistanceToOut method, if this is the case. Given the parameters of the solid it should be easily possible to check this. The simplest way to confirm this is to use the following:

2) Could you please set the navigator to a higher verbosity 
        /geometry/navigator/verbose 3
     and report on the tracking (preferably providing a file with the output from this run, rather than copying the text into a posting)

3) Could you try to run in check mode:
      - use the geometry check_mode and set verbose to 1, in case a problem is found somewhere in the geometry
       /geometry/navigator/check_mode 1
       /geometry/navigator/verbose   1

Likely this will be enough to double check that there is no obvious problem in your setup.

3) Could you experiment with the different ways of calling Locate
Instead of calling it 
    theNavigator->LocateGlobalPointAndSetup(worldPos, 0, false, true);
allow the Relative search to be "true" and see whether the result remains the same. 

  G4VPhysicalVolume* LocateGlobalPointAndSetup(const G4ThreeVector& point,
                                               const G4ThreeVector* direction=0,
                                               const G4bool pRelativeSearch=true,
                                               const G4bool ignoreDirection=true);

I will appreciate if you provide the results of these two changes. ******* Finally, just in case these are not enough to locate the problem, it could aid to:

5)  I expect that you identified this as an issue before adding the new code in ProcessHits which calls the G4Navigator.   Calling the Navigator during the event loop is fragile, especially for points outside the current volume and on surfaces.  So it will be best to double check the results which you get outside the event loop.  I suggest to try to check the location of the key points:
  ### World Position: 68.80902495440318 -191.8953954647168 54.33532473714037
  navigator volume: physiBreast 
  ### World Position: 72.85229590945139 -195.7155628014604 62.35389344721722
  navigator volume: physiSkin     <-------------

If you can simply locate these points at a different point during the run, e.g. in a EndOfTrack action, or in a separate program which does not call the Geant4 Run/Event/Tracking but just sets up the geometry, initialises and make these calls directly, it would shed light on the issue of the location of these points without any possible interference with the continuation of these tracks.

Feel free to contact me/us with partial information (e.g. from the first two points) in order to see whether it identifies the underlying issue.

Best regards, John Apostolakis

 Add Message Add Message
to: "Re: Step crosses a geometrical boundary and does not stop (from Run and Event forum)"

 Subscribe Subscribe

This site runs SLAC HyperNews version 1.11-slac-98, derived from the original HyperNews


[ Geant 4 Home | Geant 4 HyperNews | Search | Request New Forum | Feedback ]