31 JLSL: 5 Mott Insulators

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

262 Ch.

5 Mott Insulators

spins. As we discussed previously, there are three energy levels, corresponding


to total spin S = 2, 1, and 0. Setting one of these energies arbitrarily to zero,
the effective Hamiltonian has to describe two energy differences, which can be
achieved by the proper choice of the two coupling constants.
By the same token, the general form of the isotropic coupling between two
spins of size S looks like

31 = JlSl . s 2 + J2(S1 ' s2)2 + ...+ J 2 S ( S 1 a s2)2s . (5.145)

Since higher powers of (Sl .S2) arise in higher orders of perturbation theory, the
biquadratic term is the most essential correction to the standard Heisenberg
model for any S. In fact, the first experimental evidence of a biquadratic term
was found for S = 5/2 Mn++ ions (in antiferromagnetic MnO [339], as well as
in Mn-doped MgO [158]). The measured ratio Q / J -0.05 was realized to be
N

in accordance with the predictions of Anderson's theory of superexchange [20].


It is also quite compatible with our (5.144): our Q arises from the competition
of two terms of which -6t4/(U2J) is the stronger.
There is another, more formal justification for introducing (5.145). It relies
on showing that the higher powers of the scalar product (Sl . SZ)~'+I,..., are
not independent of the first 2 s operators of this kind, and therefore introducing
them would not really change the form of the Hamiltonian. This must be a
property of the algebra of the spin operators. For an illustration, let us consider
S = 1/2 where we should find that (Sl * can be expressed with S1 . S2.
Let us use Pauli matrices instead of the components of S: 2s" = uzyetc., and
exploit the wellknown property that the square of any of the Pauli matrices is
the 2 x 2 unit matrix. Then

(21 - 2 2 ) 2 = (.l"uz" + u:u; + .;u;)2 = 3 - 281 .a2 (5.146)

where we used the identities uxuY = iuz, uYuz = iuzy and uzux = iuY. In
terms of the S = 1/2 spin operators
3 1
(S1 + S2)2 = - - -s1 . s2
16 2
. (5.147)

It follows that for S = 1/2, any isotropic (pairwise) coupling can be brought to
the well-known Heisenberg form. A similar proof should be possible to justify
(5.145) for general S.

You might also like