Dear Geant4 team,

We have examined the Geant4 2.0 source codes to validate
the pair creation, and found that
the angular distribution of pair-created
e-/e+ seemed to be inappropriate.

Polar angle of generated e-/e+ should depend on the energy of e-/e+.
Actually, the Physic Reference Manual says that
Geant4 adopts the Tsai distribution, where the angular distribution
is a function of u=E*theta/mc^2
(E is the energy of e-/e+, theta is the polar angle,
m is the electron mass and c is the light speed),
whereas in the source code (G4GammaConversion.cc, line 306)
u is defined as k*theta/mc^2(k is the energy of photon).
This means that the angles of e-/e+
do not depend on the energy of generated e+/e- pairs.
We also found that the polar angles of e-/e+ are
symmetry with respect to the direction of the incident photons
(G4GammaConversion.cc, line 355),
although (we believe) these should be determined independently.

We also examined G4GammaConversion.cc of Geant4 4.0,
and confirmed that the polar angles of e-/e+ are
calculated as in Geant4 2.0.

The comparison of the angular distribution between the Geant4 source code
and the Tsai Distribution is shown in
 http://www-heaf.hepl.hiroshima-u.ac.jp/~mizuno/Ogata/PairCreation.pdf
(Figure 5.9).


Best Regards,
Tsunefumi Mizuno and Sei Ogata.