Based upon the observations (i) that their in-plane lattice constants match almost perfectly and (ii) that their electronic structures overlap in reciprocal space for one spin direction only, we predict perfect spin filtering for interfaces between graphite and (111) fcc or (0001) The observation [1,2] of giant magnetoresistance in systems where the transmission through interfaces between normal and ferromagnetic metals (FM) is spin dependent has driven a major effort to study spin filtering effects in other systems and extend applications from field sensing to storage [3], reprogrammable logic [4], and quantum computing [5]. An ideal spin filter would allow all carriers with one spin through but none with the other spin. Interfaces with half-metallic ferromagnets (HMFs) [6] should have this property but progress in exploiting it has been slow because of the difficulty of making stoichiometric HMFs with the theoretically predicted bulk properties and then making devices maintaining these properties at interfaces [7].If the nonmagnetic metal is replaced by an insulator (I) or semiconductor (SC), spin filtering still occurs giving rise to tunneling magnetoresistance (TMR) in FMjIjFM magnetic tunnel junctions and spin-injection at FMjSC interfaces. If the spin polarization of the ferromagnet is not complete, then the conductivity mismatch between metals and semiconductors or insulators has been identified as a serious obstacle to efficient spin injection [8]. It can be overcome if there is a large spin-dependent interface resistance but this is very sensitive to the detailed atomic structure and chemical composition of the interface. Knowledge of the interface structure is a necessary preliminary to analyzing spin filtering theoretically and progress has been severely hampered by the difficulty of experimentally characterizing FMjI and FMjSC interfaces.The situation improved with the confirmation of large values of TMR in tunnel barriers based upon crystalline MgO [9,10] which had been predicted by detailed electronic structure calculations [11,12]. While the record values of TMR-in excess of 500% at low temperatures [13]-are undoubtedly correlated with the crystallinity of MgO, the nature of this relationship is not trivial [14]. The sensitivity of TMR (and spin injection) to details of the interface structure [15,16] make it difficult to close the quantitative gap between theory and experiment. In view of the reactivity of the open-shell transition metal (TM) ferromagnets Fe, Co, and Ni with typical semiconductors and insulators, preparing interfaces where disorder does not dominate the spin filtering properties remains a challenge. With this in mind, we wish to draw attention to a quite different material system which should be intrinsically ordered, for which an unambiguous theoretical prediction of perfect spin filtering can be made in the absence of disorder, and which is much less sensitive to interface roughness and alloy disorder than TMR or spin injection.

We demonstrate how ab initio cluster calculations including the full Coulomb vertex can be done in the basis of the localized, generalized Wannier orbitals which describe the low-energy density functional (LDA) band structure of the infinite crystal, e.g. the transition metal 3d and oxygen 2p orbitals. The spatial extend of our 3d Wannier orbitals (orthonormalized Nth order muffin-tin orbitals) is close to that found for atomic Hartree-Fock orbitals. We define Ligand orbitals as those linear combinations of the O 2p Wannier orbitals which couple to the 3d orbitals for the chosen cluster. The use of ligand orbitals allows for a minimal Hilbert space in multiplet ligand-field theory calculations, thus reducing the computational costs substantially. The result is a fast and simple ab initio theory, which can provide useful information about local properties of correlated insulators. We compare results for NiO, MnO and SrTiO3 with x-ray absorption, inelastic x-ray scattering, and photoemission experiments. The multiplet ligand field theory parameters found by our ab initio method agree within ~10% to known experimental values

The interface-induced magnetization damping of thin ferromagnetic films in contact with normal-metal layers is calculated from first principles for clean and disordered Fe/Au and Co/Cu interfaces. Interference effects arising from coherent scattering turn out to be very small, consistent with a very small magnetic coherence length. Because the mixing conductances which govern the spin transfer are to a good approximation real-valued, the spin pumping can be described by an increased Gilbert damping factor but an unmodified gyromagnetic ratio. The results also confirm that the spin-current-induced magnetization torque is an interface effect.

Landauer's formula relates the conductance of a quantum wire or interface to transmission probabilities. Total transmission probabilities are frequently calculated using Green function techniques and an expression first derived by Caroli. Alternatively, partial transmission probabilities can be calculated from the scattering wave functions that are obtained by matching the wave functions in the scattering region to the Bloch modes of ideal bulk leads. An elegant technique for doing this, formulated originally by Ando, is here generalized to any Hamiltonian that can be represented in tight-binding form. A more compact expression for the transmission matrix elements is derived and it is shown how all the Green function results can be derived from the mode matching technique. We illustrate this for a simple model which can be studied analytically, and for an Fe|vacuum|Fe tunnel junction which we study using first-principles calculations.