a b s t r a c tWe study the possibility of ferromagnetism in metals. The metal is described by two hybridized bands one of which includes Hubbard correlation whereas the other is uncorrelated. We parametrize the ratio of the band widths and their centers as well. The original Hamiltonian is transformed into an effective and simpler one. Only one site retains the full correlation (U) while in the others acts as an internal field, the self-energy, in the framework of an alloy analogy approximation. This field, in turn, is selfconsistently determined by imposing the translational invariance of the problem. For several total electronic occupation numbers (n total ) we compare the spin dependent free energies with the corresponding paramagnetic ones. We present several results pointing out the mechanism by which the self-consistency introduces a sort of constraints, for given values of band width and band shift.& 2009 Elsevier B.V. All rights reserved.
Formulation of the problemThe origin of magnetism in itinerant ferromagnets has been object of concern in last years [1][2][3][4]. We use a two band model to describe the effect of Coulomb correlation in these metals. It consists of a Hubbard like narrow band (band a) with intrasite Coulomb interaction U, hybridized with another band, which is broad and uncorrelated (band b), through the hybridization coupling V ab . We allow a shift between the centers of the two bands and a variation of the b-band width as well, in order to describe a more general and realistic band structure. Both have strong influence on the kind of magnetic state of the metal.As in Roth's approach [8], the correlation is present in only one site (say, the origin), while in the others acts an effective spin and energy dependent but site independent field, the self-energy S s . This field is self-consistently determined by imposing the vanishing of the scattering T matrix associated to the origin, thus restoring the translational invariance of the host.Bernhard [2] addresses the correlated one-band problem, while Schwieger and Nolting [3] treats also a hybridized twoband model using the spectral density approach (SDA) and a modified alloy analogy. Their approach is equivalent to the socalled Hubbard I [5] while ours is equivalent rather to Hubbard III [6]. Our effective Hamiltonian is [7]