maximum (about Tc 0.022t) at n 0.92. The T = 0 moment is large
N N
and it may be near the Nagaoka limit35.
The strong-coupling phase boundaries can be traced by determining where the susceptibilities given in (10.80) and (10.81) diverge. The re- sult is shown by the curves A (boundary of the FM phase) and B (part of the boundary of the AFM phase) in Fig. 10.17. The lower part of the AFM phase is bounded by curve C, the result of a small-U calculation [188]. B and C do not match properly since they are derived in approxi- mations which are valid in different limits; nevertheless, the figure gives a good impression of the extent of the AFM phase. Though the bound- aries are derived by extrapolation from finite-T results, we should not hesitate to accept Fig. 10.17 as one preliminary version of the ground state phase diagram. A different version, based on QMC which is an intermediate-coupling scheme, would show AFM and incommensurate SDW phases, but no FM [118, 3831. It remains to be seen what an all-encompassing phase diagram will look like. In its gross features, the phase diagram shown in Fig. 10.17 is not unlike the variationally derived Fig. 10.10. Let us, however, point to the differences. Now we find a regime where the PM phase protrudes in between the FM and AFM phases. This has become possible because the ferromagnetic area, though horizontally almost as broad as in the Gutzwiller theory, has been pushed to higher values of U . Ferromag- netism seems to be a marginally surviving strong-coupling phenomenon whose driving force is the (projected) kinetic energy. This should be contrasted with the robustness of itinerant ferromagnetism in the fcc lattice (see Sec. 8.4 and, in particular, Fig. 8.8). Our discussion of the variational results showed that away from half- filling, the homogeneous AFM phase is likely to be unstable against phase segregation. DMFT has not yet addressed this question at strong and intermediate couplings, but there are results for the weak coupling limit where segregation occurs in an extremely small vicinity of n = 1. As a prelude, let us discuss exact half-filling. It follows from the perfect nesting property that at n = 1 for all 351n sections 8.4 and 10.6.4, we learned about some cogent arguments why the Nagaoka limit cannot be reached. Extrapolation from the finite-T results of [308] does not allow detecting a small deviation of the order of magnitude shown in (10.54).