Magnetic nanostructures on nonmagnetic or magnetic substrates have attracted strong attention due to the development of interesting experimental methods with atomic resolution. Motivated by this progress we have extended the full-potential Korringa-Kohn-Rostoker Green-function method to treat noncollinear magnetic nanostructures on surfaces. We focus on magnetic 3d impurity nanoclusters, sitting as adatoms on or in the first surface layer on Ni(001), and investigate the size and orientation of the local moments and, moreover, the stabilization of noncollinear magnetic solutions. While clusters of Fe, Co, Ni atoms are magnetically collinear, noncollinear magnetic coupling is expected for Cr and Mn clusters on surfaces of elemental ferromagnets. The origin of frustration is the competition of the antiferromagnetic exchange coupling among the Cr or Mn atoms with the antiferromagnetic (for Cr) or ferromagnetic (for Mn) exchange coupling between the impurities and the substrate. We find that Cr and Mn first-neighboring dimers and a Mn trimer on Ni(001) show noncollinear behavior nearly degenerate with the most stable collinear configuration. Increasing the distance between the dimer atoms leads to a collinear behavior, similar to the one of the single impurities. Finally, we compare some of the noncollinear ab initio results to those obtained within a classical Heisenberg model, where the exchange constants are fitted to total energies of the collinear states; the agreement is surprisingly good.

Using theoretical arguments, we show that, in order to exploit half-metallic ferromagnets in tunneling magnetoresistance (TMR) junctions, it is crucial to eliminate interface states at the Fermi level within the half-metallic gap; contrary to this, no such problem arises in giant magnetoresistance elements. Moreover, based on an a priori understanding of the electronic structure, we propose an antiferromagnetically coupled TMR element based on half-metallic zinc-blende chalcogenides, in which interface states are eliminated, as a paradigm of materials design from first principles. Our conclusions are supported by ab initio calculations.

Using the state-of-the-art screened Korringa-Kohn-Rostoker Green function method we study the electronic and magnetic properties of NiMnSb and similar Heusler alloys. We show that all these compounds are half-metals, e.g. the minority-spin band is semiconducting and the Fermi level falls within this gap resulting in 100% spin polarization at the Fermi level. The total spin moment M-t shows the so-called Slater-Pauling behaviour and scales with the total valence charge Z(t) following the rule M-t = Z(t) - 18 for half and M-t = Z(t) - 24 for full Heusler alloys. These rules are connected to the origin of the gap. Finally we show that the inclusion of the spin-orbit interaction in our calculations kills the half-metallic gap but the spin-polarization at the Fermi level can be still very high, similar to 99% for NiMnSb, but much lower for a half-metallic compound like zinc-blende MnBi (77%).