Factorization is the fundamental concept in order to apply
QCD calculations to description of hadronic reactions. According to Fac-
torization, theoretical study of any hadronic process includes both parton
scattering and parton distributions. The partonic sub-processes are calcu-
lated with the use of regular methods of perturbative QCD. In contrast, the
parton distributions are introduced purely phenomenologically, without any
theoretical grounds. There are two popular kinds of Factorization in the
literature: Collinear and kT - factorizations. They were introduced indepen-
dently of each other. We show that both the kT - and collinear factorization
can be obtained by consecutive reductions of some more general (Basic) fac-
torization. Each of these reductions is an approximation valid under certain
assumptions.
First, the transitions from Basic to kT - factorization assumes that the mo-
menta of the partons connecting the perturbative and non-perturbative blobs
are mostly transverse. This assumption fairly agrees both with the DGLAP
and BFKL.
Second, if the unintegrated parton distributions in kT -factorization have a
maximum(s) in kT , then kT - factorization can be reduced to collinear factor-
ization. The sharper the maximum is, the better is accuracy of the transition.
This assumption can be checked with analysis of available experimental data.
Integration over momenta of the connecting partons in the Basic factorization
for amplitudes of the forward Compton scattering o hadrons must yield a
nite result. This obvious requirement allowed us to deduce theoretical con-
straints on the parton distributions both in kT - and Collinear factorizations.