We have introduced hydrogen diffusion in our evolution code in such a
way that the chemical composition is computed taking simultaneously in
to account diffusion and nuclear reactions. The first model was obtain
ed with the diffusion coefficients given by Thoul et al.(1995). Its fr
equencies show that the model should be improved. Five other models be
en have obtained with modifications of the diffusion coefficients and/
or an increase of the opacity below the convective envelope. Several o
f these models show definite improvements such that most of the remain
ing sources of discrepancies are located in the convective envelope an
d mainly in its outermost layers where the superadiabatic gradient is
large. The models have also been compared with Basu's seismic model. I
n the domain of confidence of the latter, the relative differences in
sound speed is less than 0.1% for models which fit best the observed f
requencies.