Vous êtes sur la page 1sur 1

5.

5 Correlated Insulators 239

It is a different question that at finite temperatures, we cannot really


tell a (poor) metal from an insulator: they both have finite conductivi-
ties and need not differ in their symmetry properties. The change from
PI to PM could be perfectly continuous, and in fact, at high enough
temperatures it is just that. The PM-PI phase boundary in Fig. 5.5 is
a line of first-order transitions, and it ends at a critical point. It reminds
us of the well-known liquid-gas phase boundary. In both ewes, the tran-
sition occurs without a change in global symmetries, being associated
with a jump in density-like quantities, and in the entropy.
We can even think of a reason why there should be a first-order tran-
sition from a low-temperature correlated PM phase to a high-tempera-
ture PI phase. The conduction electron entropy of a met& is 5' N

JCBTITF, where T ' is the Fermi temperature. On the other hand, the
electrons are localized in the paramagnetic insulator, and therefore they
have the spin disorder entropy S N kg ln2. Making the plausible as-
sumption that In 2 >> T/TF,we predict that at some finite temperature,
the free energy 7 = U - TS of the insulator becomes lower than that
of the metal. This explains why the PM-PI phase boundary in Fig.
5.5 is slanting to the right16. We mention that the first-order PM-PI
transition is also called a Mott transition, and rightly: it takes from a
metallic state to an insulator in which the electrons are collectively l e
calized. Some people would say that only this is a true Mott transition,
since it is not associated with magnetic ordering.
The schematic phase diagram shown in the right-hand panel of
Fig. 5.5 sums up our expectations about the typical behaviour of a
half-filled non-degenerate Hubbard modelI7. Essentially the same phase
diagram is obtained from studies on non-nesting variants of the infinite
that it keeps some of its Mott-character. SDW's are discussed in Sec. 7.6.
lSThis feature makes it possible that at certain pressures (or Cr concentrations)
two consecutive transitions can be observed: first from AFI to PM, and at a higher
T, from P M to PI (see the curve z = 0.01 in Fig. 5.4, right).
l'"Typicd" implies that the band structure does not have the perfect nesting
property which would require a fine-tuning of the parameters. The schematic phase

-
diagram for a nested F'ermi surface is shown in Fig. 7.6. Furthermore, note that since
UIW is increasing to the left, the antiferromagnetic exchange coupling J W a / U
should be decreasing to the left. That's why we were careful to draw a gently sloping
AFI-PI boundary in the schematic phase diagram. This feature does not become
apparent in the experimental phase boundary.

Vous aimerez peut-être aussi