7.

7 l’kansitionMetals and Alloys 387

the 3d and 43 bands overlap, they both have non-integer fillings. For in-
stance, it is reckoned that in paramagnetic Fe, the 3d band holds 6.5 N

electrons, while the 43 band has N 1.5 electrons. Furthermore, the ef-
fective Hubbard U is diminished because the light 3-electrons screen the
electron-elec tron interact ion effectively.
Introducing lattice models of interacting fermions is the standard
approach in this book but at the present, it is certainly not the most
promising way of thinking about transition metals. Another possibility
is to refer to the results of up-to-date electronic structure calculations.
Density functional theory [92], or more specifically: the application of
the local spin density approximation (LSDA) in band theory, has been
remarkably successful in explaining the properties of TM elements, and
their alloys [15]. We discuss the relationship of LSDA to lattice fermion
models in Sec. 7.7.2. In spite of its many successes in describing weakly
or moderately correlated systems, LSDA is not a proper theory of elec-
tron correlations; for instance, it often fails to find the Mott-Hubbard
gap of Mott insulators. However, it works very well for the ferromag-
netic iron group elements. The total energy given by LSDA can (with
some further approximations) be brought to the form (7.26), confirming
the picture of exchange-split bands. In principle we expect the energy
difference &f-PkJ- to be k-dependent but the finding [141] that the split-
ting is, in fact, more-or-less constant allows to use the parametrization
(7.27), and to calculate the Stoner parameter I from first principle^^^.
I is found to lie in the range O.7eV-lev for the 3d series (increasing
towards Ni). Somewhat smaller values are found for the 4d and 5d se-
ries. This shows that the electron-electron interaction is of intermediate
strength.
In view of the relative constancy of I , going from element to element
in the 3d series amounts, roughly speaking, to changing the d-band
filling. Varied behaviour is found [442]. Ti and V are paramagnetic
metals. Cr has an interesting spin density wave phase which we discuss

32TheLSDA calculations treat the true degenerate d-band which is hybridized with
a wide s-band, and the interaction is the full Coulomb interaction. Thus the Stoner
I should be regarded as an effective coupling constant which no longer has the simple
meaning of the Hubbard U [357]. I N l e v may well imply a Hubbard U of several
eV, which confirms that we have to do with an intermediate coupling problem.