[Jlab-scicomp-briefs] Geant4-9.6.0 has major bug in magnetic field tracking routines

[Brad Sawatzky]

Just a note that Geant4-9.6.0 has a regression in the stepping code that
can result in incorrect particle motion through a magnetic field.  Here's an
excerpt from http://geant4.web.cern.ch/geant4/support/Patch4.9.6-1.txt
  Fixing major bug of migration of multiple-scattering sampling from
  PostStepDoIt() to AlongStepDoIt(), responsible for observed mis-
  behavior in presence of magnetic-field.

I didn't dig into the details of the fix, but there is some subtlety to
how this bug manifests.  Particle tracking in a B-field breaks
completely in the Qweak simulation (particles don't seem to bend at
all), but I have other simulations where it seems to work correctly
(qualitatively, at least).

Geant4-9.6.0 was released in Nov 2012 and introduced this bug.
Geant4-9.6.1 was released in Feb 2013 and has the fix.

I see that Maurizio *does* have Geant4-9.6.1 available in
/site/12gev_phys/.  Given the significance of the bug, I would
recommend that Geant4-9.6.0 be deleted (perhaps with a notice).

-- Brad