establishing connection with our previous discussion of CeBc, URuzSi2,
etc. However, we do not follow this line, because it would require a lengthy formal development. Let us recall that in Sec. 3.6 we found a different reason for the aris- ing of an orbital superlattice: the cooperative Jahn-Teller effect. The Jahn-Teller effect does not necessarily “want” the same kind of order as the kinetic exchange processes described here. The resulting com- petit ion bet ween the electron-elec t ron and electron-lat t ice interact iong gives rise to complicated physics. As for LaMnOs, the resulting or- bital order can be described approximately as the in-plane alternation of d322-7.2 and d3y2-T2 orbitals (Fig. 5.2, right). It is interesting to notice that there is no alternation along the c axis, corresponding to the fact that in that direction, the spin order alternates (compare to Fig. 5.3, right). We explain this finding shortly. To describe orbital ordering, it is useful to introduce a pseudospin variable: let rj” = 1/2 if cpa is occupied on site j , while T; = -1/2 belongs to the occupation of Vb. Process I1 did not change the spin but it did flip the pseudospin; it is described by
N -
In contrast, Process I belongs to
u a (7-f +f) (Ti + f) s&.
Many terms like these have to be collected. The full effective Hamilto-
couplings -7172, as well as mixed terms of the sort -
nian contains, in addition to the usual spin-spin coupling, also orbital (S152)(r1~2). Fig. 5.2 shows the alternating orbital order which we would expect to be accompanied by parallel spin alignment. It is interesting that a mean field calculation based on the assumption of both spin and orbital ordering finds that the orbital superlattice may survive to higher tem- peratures than the spin magnetism [76].This case seeems to be realized in Lao.5Sr1.5Mn04 [301] (see p 480). ‘Here we may lump together all the lattice strain-electricmultipole and multipole- multipole interaction terms which do not contain the magnetic degrees of freedom explicitly.