Why RKKY exchange integrals are inappropriate to describe ferromagnetism in diluted magnetic semiconductors

We calculate Curie temperatures and study the stability of ferromagnetism in diluted magnetic materials, taking as a model for the exchange between magnetic impurities a damped Ruderman-Kittel-Kasuya-Yosida (RKKY) interaction and a short-range term representing the effects of superexchange. To properly include effects of spin and thermal fluctuations as well as geometric disorder, we solve the effective Heisenberg Hamiltonian by means of a recently developed semi-analytical approach. This approach, "self-consistent local Random Phase Approximation (SC-LRPA)," is explained. We show that previous mean-field treatments, which have been widely used in the literature, largely overestimate both the Curie temperatures and the stability of ferromagnetism as a function of carrier density. The discrepancy when compared to the current approach was that the effects of frustration in RKKY oscillations had been strongly underestimated by such simple mean-field theories. We argue that the use, as is frequent, of a weakly-disordered RKKY exchange to model ferromagnetism in diluted III-V systems, is inconsistent with the observation of ferromagnetism over a wide region of itinerant carrier densities. This may be puzzling when compared to the apparent success of calculations based on ab initio estimates of the coupling; we propose a resolution to this issue by taking RKKY-like interactions between resonant states close to the Fermi level.