The physical contributions to the KNiF3 magnetic exchange coupling integral have been obtained from specially designed ab initio cluster model calculations. Three important mechanisms have been identified. These are the derealization of the magnetic orbitals into the anion "p" band, the variational contribution of the second-order interactions, and the many-body terms "hidden" in the two-body operator of the Heisenberg Hamiltonian.PACS numbers: 75.10.Jm, 71.10.+x Magnetic properties of ionic solids have experienced a renewed interest due to the discovery of high-7V superconductors [1,2]. A relationship between magnetism and superconductivity arises from the new superconductor mother compounds, such as La2CuC>4, which are antiferromagnetic insulators close to the Mott transition [3]. From the point of view of theory, two approaches have been used. The first one is based on local spin-density calculations [4] whereas the second one uses simplified model Hamiltonians [5][6][7]. While highly desirable, a fully ab initio description has not been reported yet. The reason for this lack is that, in spite of the high degree of accuracy that can be achieved by the modern quantum chemistry, the problem of magnetism in solids is very difficult since it involves very small energy differences.For molecular complexes, de Loth et al.[8] suggested a second-order nonempirical treatment based on the theory of effective Hamiltonians. A variational version of this procedure has been reported recently and generalized to different physical situations [9,10]. In principle, there is no reason why such theoretical approaches could not be applied to solids as well if a suitable model is used. This is the aim of the present work.In a series of papers Malrieu and co-workers have used a magnetic model to study alkali metals, conjugated hydrocarbons, and related systems [11][12][13][14]. In this model the exchange coupling integral C/) is extracted from accurate calculations in simple models. The idea behind this approach is very attractive because it implies that the magnetic behavior of extended systems can be obtained from nonempirical model Hamiltonians with parameters extracted from accurate quantum chemical calculations on small models, usually dimers.The aim of this work is to explore the magnetic structure of a simple ionic solid such as KNiF3 using a parameter-free approach (see also Refs. [15,16]). We will use different cluster models for KNiF 3 and different ab initio wave functions to show that there are three main factors governing magnetism in KNiF3. For the first time, we will quantify the importance of these contributions to the final coupling exchange constant. The KNiF3 cubic perovskite is known to be an excellent example of a simple nearest-neighbor exchange Heisenberg system. The spin Hamiltonian can be written simply as
where (ij) means that the summation in (1) is over all nearest-neighbor pairs / and j Ni 2+ cations on KNiF3 structure. Moreover, the ground state of Ni 2+ in KNiF3 is 3 F so each cation can be simply se...