Download as pdf or txt
Download as pdf or txt
You are on page 1of 1

462 Ch.

8 Ferromagnetism in Hubbard Models

We do not attempt to review the results on the extended one-band


Hubbard model. Let us, however, quote the large-U effective exchange
coupling found in the solution to Problem 5.5 (see (5.108) on p. 254)
4(t - X ) 2
Jeff = -2F. (8.35)
u-v
Jeff tends to change sign as U is increased. Thus though the large
Hubbard U in itself may not make the system ferromagnetic, it gives
rise to a situation where even a quite weak direct exchange F is sufficient
to turn the system into a ferromagnet (see [437] for a detailed study).
What is the upshot of all these arguments: did the study of the
single-orbital Hubbard model enable us to understand itinerant ferro-
magnetism, or did it not? It is possible to take the point of view that, to
some extent, it did. We may declare ourselves content with the available
evidence that the fcc lattice shows robust ferromagnetism, and accept
it as a semi-quantitative explanation for the properties of Ni. As for
Co, its lattice has the hcp structure, and that must be just as inclined
towards ferromagnetism as the fcc lattice3'. This still leaves us with
the ferromagnetism of Fe to account for. Though there is a widely used
definition according to which iron is a weaker ferromagnet31 than Co
and Ni, and one could say that this is in agreement with the fact the
bipartite bcc lattice shows only a weak tendency to ferromagnetism, we
prefer to take a more balanced view. A Curie temperature Tc = 1044K
does not suggest anything weak, and the value ppar/psat -
1 also shows
that Fe is as far from being a weak itinerant ferromagnet as a system
can be (see Table 7.1, p. 400). There is really no need to hurry and try
to explain everything in terms of the single-orbital model, since allow-
ing for the degeneracy of the 3d band, and the concomitant intraatomic
exchange, adds further important features to our picture of itinerant
ferromagnet ism.
301t would take some arguing to show this. With nearest-neighbour hopping, the
fcc and hcp lattices have exactly the same DOS, though the band structures are
different. It is only at the SKA level, that a band is sufficiently characterized by
its DOS; more complicated states are sensitive to the details of the band structure.
Furthermore, it turned out that the relevant fcc model is a t 1 - h model, and the
corresponding generalized hcp models have not been studied yet.
311nthe sense that its T = 0 magnetization is decidedly smaller than the saturation
value allowed by the band filling.

You might also like