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

500 Ch.

9 The GutzwiJJer Variational Method

According to the estimate (9.3), for U > 41E1 the ground state energy
would become positive! It would be absurd to accept this conclusion
since, for large U’s, any of the neutral states I$neut) such as the one
shown in the right-hand picture of Fig. 9.1, would be a better trial
state, giving exactly 0 energy (the Wannier states have 0 band energy
and there is no double occupation). It means that for sufficiently large
U s , we must arrive at a Heitler-London type description of the solid.
This was our naive argument for a Mott-transition in Sec. 4.2. Here we
attempt to be less naive. Instead of letting nd jump from its uncorrelated
value 1/4 to zero, we allow for the obvious fact that, as soon as we switch
on the Hubbard U,there must be a tendency to reduce the density of
doubly occupied sites.
Let us return to Fig. 9.l.c. Note that such a fully neutral state is
expected only at U -+ 00. At any finite U , hopping will inevitably mix
in the charged states by the process

(9.4)
so the optimally chosen ground state must be something in between
the extremes shown in Fig. 9.1.a and Fig. 9.l.q with the concentration
of doubly occupied sites

lying at some intermediate value


1
O<n(j<-.
4
As U is increased from 0 towards 00, nd must decrease from the un-
correlated value 1/4 to 0. We want to find out how nd depends on
u/w.
Remember that we found one possible answer to this question in
our spin density wave mean field theory: according to eqn. (7.89),
nd = 1/4-m2 ( m being the sublattice magnetization) which indeed per-
forms a smooth interpolation between the aforementioned limits. The

You might also like