We have used realistic Comptonization models to fit high-quality BeppoSAX d
ata of six Seyfert galaxies. Our main effort was to adopt a Comptonization
model taking into account the anisotropy of the soft photon field. The most
important consequence is a reduction of the first scattering order, which
produces a break (the so-called anisotropy break) in the outgoing spectra.
Thus, anisotropic Comptonization models yield spectra with convex curvature
s. The physical parameters of the hot corona (i.e., the temperature and opt
ical depth) obtained by fitting this class of models to broadband X-ray spe
ctra are substantially different from those derived by fitting the same dat
a with the power law + cutoff model commonly used in the literature. In par
ticular, our best fits with Comptonization models in slab geometry give a t
emperature generally much larger and an optical depth much smaller than tho
se derived from the power law + cutoff fits, using standard Comptonization
formulae. The estimate of the reflection normalization is also larger with
the slab corona model. For most objects of our sample, both models give Com
pton parameter values larger than expected in a slab corona geometry, sugge
sting a more "photon-starved" X-ray source configuration. Finally, the two
models provide different trends and correlation between the physical parame
ters: for instance, with the power law + cutoff fits, we obtain a correlati
on between the reflection normalization and the corona temperature, whereas
we find an anticorrelation between these parameters with the slab geometry
. These differences have major consequences for the physical interpretation
of the data. In the framework of reprocessing models, the cutoff power-law
best-fit results suggest that the thermal corona is dominated by electron-
positron pairs. On the contrary, the slab corona model is in better agreeme
nt with a low pair density solution.