There is a serious flaw in the Geant4.5.1 procedure that randomly
choses the target atom for a hadronic interaction.  The net result of
this error is that the last element stored in the element vector is
preferentially selected, regardless of its relative abundance.  In
other words, the material definition implemented in the detector
construction procedure is essentially meaningless for hadronic
processes.

The problem lies in the subroutine G4HadronicProcess::ChooseAandZ()
(in the file G4HadronicProcess.cc), and specifically in the code that
assimilates data in the vector runningSum[].  As programmed is version
4.5.1 the value of runningSum[i] is the product of density and
microscopic cross section for element i.  However, it should be the
sum of these quantities for elements 0 to i.

A patch for the code from G4HadronicProcess.cc starting at line 62 is
reproduced below with the original code commented.

    for( i=0; i < numberOfElements; ++i )
    {
      //   runningSum.push_back(theAtomicNumberDensity[i] *
      //     dispatch->GetMicroscopicCrossSection( aParticle,
(*theElementVector)[i], aTemp));
      //   crossSectionTotal+=runningSum[i];
      crossSectionTotal += theAtomicNumberDensity[i] *
        dispatch->GetMicroscopicCrossSection( aParticle, (*theElementVector)[i],
aTemp);
      runningSum.push_back(crossSectionTotal);
    }