Professional Documents
Culture Documents
The direct exchange coupling F (and with it, Y)looks small in any
comparison, so we might consider neglecting at least %F and % y in
“reasonable” extended Hubbard models. There is, however, a convinc-
ing counter-argument against omitting 3 c ~ .The limit of strong-coupling
Mott insulators is a physically relevant case. In zeroth order, all the
neutral configurations have the same energy contribution from the di-
agonal Coulomb terms, and all hopping processes are blocked. The ex-
change term Rflpsplits the degeneracy, favouring parallel spin alignment.
Naturally, we understand that this cannot be the full story, otherwise
Mott insulators would be ferromagnetic! The solution to Problem 5.5
(see p. 254) should convince us that the effective exchange coupling is
Jeff N -F i-f t - X ) 2 / V . Thus F should not be compared to the large
energies U and V , but rather to the perturbatively small (t - X ) 2 / V .
In this context, there is no a priori reason to neglect 3 c ~ .
We do not attempt even to sketch the phase diagram of the extended
model X,~,,apart from reconsidering the question of ferromagnetism
in Ch. 8. Let us, however, make one general remark. We learned
in Sec. 4.5.2 that the phase diagram of the simple Hubbard model is
symmetrical about half-filling for bipartite lattices. What about the
extended model? We find that the argument using the electron-hole
transformation (4.45)goes through for most of the new terms, but not
for Rftx. The assisted hopping term (5.68) breaks the electron-hole
symmetry even for bipartite lattices, and the resulting phase diagram
must be asymmetric.