We present a. study of the superconducting properties of models containing
the Hubbard rnpulsion term. This strnng on-site repulsion
is considered as a key ingTedient of the high tempera.turn superconductivity.
Though the fact that for the normal, lovv temperature
superconductors, rnpulsion destroys superconducting order, it is argued
in the present thesis, that for the pairing of the d-·wave symmetry
in the strongly correlated electronic systems its effect is to
enhance and may be to cause superconductivity. Various methods
such as Variational Monte Carlo, Gutzwiller Approximation, Time
Dependent Hartree-Fock and Fixed Node Approximation have been
used to investigate t - U - W model, t - U - J - V and pure Hubbard
models.
In this thesis, by considering correlations contribution to the BCS
condensation energy due to the Hubbard U it is shown that the latter
lowers the total energy of a d-wave superconductor in the weak
coupling limit, thus enforcing the stability of such superconductor.
This effect appears to be mainly due to the enhancement of the
spin fluctuations near the nesting vector Q = ( π, π).
It is then studied the crossover from weak to strong coupling
regimes in t - U - J - V model by increasing U. Remarkably in this
model an order of magnitude growth of the superconducting order
parameter is found and explained as being due to the Hubbard repulsion.
vVe fi.nd also, that the pairings, originally induced by spin
or charge fluctuations upon increase of U are differently renormalized,
being the former enhanced, while the latter suppressed.
In the fi.nal part of the thesis the superconductivity in the pure
Hubbard model is carefully studied by means of Variational Monte
Carlo and related numerical methods aimed to improve variational
results. VVe observe the onset of strong coupling superconductivity
at U/t ~ 7 within the Fixed Node Approximation in the systems of
large size and compare our results for small clusters with those of
Lanczos diagonalization. We show that Variational Monte Carlo,
though overestimating the quasiparticle weight (ZvMc > Zexact)
succeeds in reproducing the correct pairing between quasiparticles.
