When a material definition contains one or more element with a fraction of zero, the definition is accepted by GEANT4 (no error, no warning); however simulation results are erroneous. We are reporting this problem now (as the Gate collaboration) but it was initially mentioned in 2006 in the hypernews (see URL above). The problem is reported using low-energy processes; it is unknown at this time if other processes are also involved. Whether a zero fraction should be allowed in a material definition is debatable. If it is not, then an error message should be issued; if it is, then of course valid results should follow.
Old problem. I remember at the time it was submitted that the situation was uneasy in the LowE group. After discussion with Vladimir and Michel, it looked to be a problem in the LowE package (suspicion was kind of division by the element fraction, regardless this fraction could be zero). This is my recollection of the situation.
Problem fixed thanks to the new design of low energy classes
