We present ab initio calculations for orbital moments and anisotropy energies of 3d and 5d adatoms on the Ag(001) surface, based on density functional theory, including Brooks' orbital polarization (OP) term, and applying a fully relativistic Korringa-Kohn-Rostoker -Green's function method. In general, we find unusually large orbital moments and anisotropy energies, e.g., in the 3d series, 2.57m B and 174 meV for Co, and, in the 5d series, 1.78m B and 142 meV for Os. These magnetic properties are determined mainly by the OP and even exist without spin-orbit coupling. DOI: 10.1103/PhysRevLett.86.2146 The interest in surface magnetism is primarily caused by the enhancement of the spin moments at surfaces being driven by the reduced coordination. A typical example for this effect is iron, for which the bulk moment (2.15m B ) is enhanced at the (001) [6]. Here the 4d and 5d atoms, being nonmagnetic as impurities in the bulk, show as adatoms very large local moments comparable with the free atom values. Sizable moments also survive when these atoms are incorporated into the first layer.In contrast to the spin magnetism, the orbital magnetism in solids has its origin in the spin-orbit interaction and is closely connected with the magnetocrystalline anisotropy, with magneto-optical effects and magnetic x-ray dichroism. In this paper we address the enhancement of orbital moments at surfaces. It is well known that the orbital moments are "quenched" in the bulk, i.e., strongly suppressed by the crystal field splitting, the strong hybridization with the neighboring atoms, respectively. Thus, calculations yield very small orbital moments: 0.049m B , 0.075m B , and 0.042m B for bcc Fe, hcp Co, and fcc Ni [0.082m B , 0.123m B , and 0.058m B if Brooks' orbital polarization (OP) is included] [7]. At surfaces ab initio calculations show that also the orbital moments are enhanced, to e.g., a value of 0.090m B at the hcp Co(0001) surface (0.158m B if Brooks' OP is included) [7,8]. Even larger orbital moments are obtained for the 3d monolayers, e.g., 0.121m B for a Co monolayer on Cu(100) (0.261m B including OP) [7,9]. Thus, at surfaces the quenching of the orbital moments is less pronounced due to the reduced hybridization. However, it is important to realize that these enhanced orbital moments are still an order of magnitude smaller than the corresponding free atom values, as given by Hund's second rule. Thus, the orbital moments are to a large extent also quenched at the surface, as it seems to be the general rule in metallic environments.However, this rule can have exceptions. Riegel and co-workers [10] have already shown that Fe impurities, being injected into alkali metals, show hyperfine properties which indicate very large orbital moments, probably the full atomic values. The low and more or less constant electron density of the alkali hosts is responsible for this behavior. Here we will predict by density functional calculations that single 3d and 5d transition metal impurities on the Ag(001) surface can have very large or...

We present ab initio calculations of the orbital moments and magnetocrystalline anisotropy energies for 3d, 4d, and 5d transition-metal adatoms and for some selected small clusters on the ͑001͒ surfaces of Ag and Au. The calculations are based on the local density approximation of density functional theory and apply a fully relativistic Koringa-Kohn-Rostoker Green's function method. Due to the reduced coordination of the adatoms and the weak hybridization with the substrate, we find fairly large orbital moments, in particular for the elements towards the end of the series. The general trend can be understood from a simple tight-binding model. The orbital moments are connected with very large anisotropy energies. While the orbital moments are on the Ag substrate somewhat larger than on Au, the magnetic anisotropy has about the same size for both substrates. Calculations for small clusters of Fe, Ru, and Os adatoms show, that due to interaction effects the orbital moments are strongly reduced, e.g., by 50% for the dimer atoms. The size of the reduction correlates well with the coordination number. Similar reductions also occur for the magnetic anisotropy energies.

Motivated by recent STM results for Cr on Fe(001) we present a detailed ab-initio study of the alloying process in the dilute limit for 3d atoms on the Fe(001) surface. The calculations are based on local density functional theory and apply a KKR{Green's function method for impurities on surfaces. For practically all 3d transition metal impurities on Fe(001) we nd a strong tendency for a direct exchange mechanism into the rst surface layer. The early 3d impurities V, Cr and Mn strongly repel each other on neighboring positions within the rst layer, while Ni and Cu atoms show a moderate repulsion. The ab{initio results are in good agreement with STM studies for Cr/Fe(001) by the NIST group and present valuable predictions for all 3d/Fe(001) systems.

We present ab initio calculations of scanning tunneling spectra for the Fe͑001͒ surface and for 3d impurities in this surface. The calculations are performed by the full-potential Korringa-Kohn-Rostoker Green's-function method, and also partly by the full-potential linearized augmented-plane-wave method. For the clean Fe͑001͒ surface we demonstrate that the correct tunneling spectrum is only obtained in a full potential treatment, while the atomic-sphere approximation yields incorrect results. For 3d impurities in the surface layer, peaks appear in the spectra due to surfacelike states localized on the impurity site. Our results can explain recent scanning tunneling microscopy experiments on Cr impurities in the Fe͑001͒ surface, and predict that chemical identification is also possible for many other transition-metal impurities.

No abstract