7.

7 Tkansition Metals and Alloys 393

‘electron glue’ is such a complex many-body system that it cannot be

usefully parametrized” [144] - not, if we are so ambitious as to require
a unified description of low- and high-energy phenomena, ranging from
O.OleV to 100eV.
Though the above statement does not exclude the possibility of find-
ing a relatively simple low-energy effective Hamiltonian, the fact remains
that it is quite difficult to establish the relationship between an LSDA
calculation, and the study of a model Hamiltonian with a small local
basis, and a few interaction parameters. Let us recall that the Hubbard
model was constructed with a single local orbital, and a single inter-
action parameter! One studies such, admittedly oversimplified, models
in the hope that they allow the identification of the “driving force” of
certain types of ordering. The approach is useful only if we have some
reason to believe that, after including many more terms into the Hamil-
tonian, it becomes realistic enough to describe a system, but the nature
of ordering is still basically the same as in the simple model. Assum-
ing that this is the case, the understanding gained by working with the
simple model is of great value.
The problem of the connection between LSDA and Hubbard-model
physics has been brought into the focus of attention by the reviving
controversy about itinerant ferromagnetism. LSDA has been able to
describe many properties of the ferromagnetic iron group elements in
impressive detail (Sec. 7.7.4). On the other hand, it is not clear how
to account for band ferromagnetism starting from the Hubbard model
(Ch. 8). Thus the question can be posed: which is the simplest lattice
fermion model which would have the same qualitative features as the
LSDA description of iron?
We return to a detailed discussion of this question in Ch. 8. There
are indications that even the one-band Hubbard model is capable of
ferromagnetic ordering in a physically reasonable parameter range if we
“fine-tune” the kinetic energy by choosing certain non-bipartite lattices
or/and farther-neighbour hopping. Another relevant extension is to
degenerate models37, or to many-band models. Time will show which
should be derived from the underlying Coulomb force. LSDA has no adjustable
interaction parameters.
37These allow to bring into play the intraatomic exhange (responsible for Hund’s