Sie sind auf Seite 1von 1

Solutions to the Problems 407

chemical potential remains pinned at p = 0 for arbitrary temperatures. At


high T >> W , expanding in terms of WIT gives

where 12 is the second moment of the DOS. The leading term is Curie-like
and corresponds to half of the sites having an unpaired electron. We were
discussing (Sec. 4.2) that in a half-filled band, on the average, half of the
sites are occupied by one electron each. At low T , the rapid hopping mo-
tion mixes the configurations and quenches most of the spins. At high T ,
the quantum-mechanical coherence between the configurations is destroyed by
thermal fluctuations, and the spins “become visible”. This is of mere aca-
demic interest for ordinary metals. However, for correlation-induced effective
narrow bands of widths W 5 kBTroom,the crossover from a low-temperature
Pauli susceptibility to a high-temperature Curie susceptibility is an important
effect.
Let us note that the assumption about a symmetrical DOS was made
merely for convenience: we did not have to worry about the temperature de-
pendence of p. The result (7.109) holds for an arbitrary form of p(e).
c) According to (7.108), the necessary condition for the spin susceptibility
to increase as T is raised from T = 0 is that p(c) must have a sufficiently
large positive curvature at c = CF. It is certainly satisfied if we place CF at the
bottom of the central valley of the double-peak DOS shown in the left panel
of Fig. 7.12. We choose

(7.1 10)

and assume that the band is half-filled, so that EF = 0, and p = 0 at all


T. The temperature dependent spin susceptibility (7.107) of non-interacting
band electrons is shown (in units of (gpB)2)as the lower curve in the right-hand
panel of Fig. 7.12.
The effect of switching on U is described by (7.44). We will learn from
solving Problem 7.2 that the T = 0 Stoner instability sets in at U w 4.23, thus
U = 1.8is for this problem a modest interaction. Still, it leads to a much more
pronounced maximum in x(T) (Fig. 7.12, right, upper curve). The reason is
that the enhancement factor also grows with increasing x ( O ) (2’).
The temperature dependence of the measured susceptibility of Pd (and
even more spectacularly Y(Co.saAl.ll)z [344]) shows a maximum (Fig. 7.7)
which is reminiscent of the feature found by us. We should not jump to the
conclusion that EF in Pd is sitting in a local minimum as in our model cal-
culation. The necessary condition for the existence of a maximum is merely

Das könnte Ihnen auch gefallen