| Popis: |
The description of the magnetic properties of interacting many-particle systems has been one of the most important goals of physics. The problem is to derive the magnetic properties of such systems from quantum mechanical principles. It is well understood that the magnetization in an atom described by quantum numbers, spin (S), orbital (L), and total angular momentum (J) of its electrons. A set of guidelines, known as Hund's rules, discovered by Friedrich Hermann Hunds help us to determine the quantum numbers for the ground states of free atoms. The question ``to which extent are Hund's rules applicable on different systems such as molecules and solids?'' is still on the agenda. The main problem is that of finding the ground state of the considered system. Density functional theory (DFT) methods apparently are the most widely spread self-consistent methods to investigate the ground state properties. This is due to their high computational efficiency and very good accuracy. In the framework of DFT, usually the total energy is decomposed into kinetic energy, Coulomb energy, and a term called the exchange-correlation energy. Taking into account the relativistic kinetic energy leads to direct and indirect relativistic effects on the electronic structure of a solid. The most pronounced direct effect (although not the biggest in magnitude) is the spin-orbit splitting of band states. A well-known indirect relativistic effect is the change of screening of valence electrons from the nuclear charge by inner-shell electrons. One can ask that how relativistic effects come into play in ordinary density functional theory. Of course ordinary density functional theory does not include those effect. Four-current density functional theory (CDFT), the quantum electrodynamic version of the Hohenberg-Kohn theory is a powerful tool to treat relativistic effects. Although it is principally designed for systems in strong magnetic fields, CDFT can also be applied in situations where currents are present without external magnetic fields. As already pointed out by Rajagopal and Callaway (1973), the most natural way to incorporate magnetism into DFT is the generalization to CDFT. These authors, however, treated its most simple approximation, the spin density functional theory (SDFT), which keeps the spin current only and neglects completely correlation effects of orbital currents. By using the Kohn-Sham-Dirac (KSD) equation, spin-orbit coupling is introduced kinematically. The part of the orbital magnetism that is a consequence of Hund's second rule coupling is absent in this theory and there is not any more a one-to-one mapping of spin densities onto external fields. In solids, in particular in metals, the importance of Hund's second rule coupling (orbital polarization) and Hund's third rule (spin-orbit coupling) is usually interchanged in comparison to atoms. Thus, in applications of the relativistic CDFT to solids, the usual way has been to keep the spin-orbit coupling in the KSD equation (an extension to ordinary Kohn-Sham (KS) equation) and to neglect the orbital contribution to the total current density and approximate exchange-correlation energy functional with spin density only. This scheme includes a spontaneous exchange and correlation spin polarization. Orbital polarization, on the other hand, comes into play not as a correlation effect but also as an effect due to the interplay of spin polarization and spin-orbit coupling: In the presence of both couplings, time reversal symmetry is broken and a non-zero orbital current density may occur. Application of this scheme to 3d and 4f magnets yields orbital moments that are smaller than related experimental values by typically a factor of two. Orbital magnetism in a solid is strongly influenced by the ligand field, originating from the structural environment and geometry of the solid. The orbital moments in a solid with cubic symmetry are expected to be quenched if spin-orbit coupling is neglected. However, spin-orbit coupling induces orbital moments, accordingly. The relativistic nature of the spin-orbit coupling requires orbital magnetism to be treated within QED, and the treatment of QED in solids is possible in the frame of current density functional theory. The kinematic spin-orbit coupling is accounted for in many DFT calculations of magnetic systems within the LSDA. However, a strong deviation of the LSDA orbital moments from experiment is found in such approaches. To avoid such deviations, orbital polarization corrections would be desirable. In this Thesis, those corrections have been investigated in the framework of CDFT. After a short review for CDFT in Chapter 2, in Chapter 3, an "ad hoc" OP correction term (OPB) suggested by Brooks and Eriksson is given. This correction in some cases gives quite reasonable corrections to orbital moments of magnetic materials. Another OP correction (OPE), which has been introduced recently, was derived from the CDFT in the non-relativistic limit. Unfortunately, the program can only incompletely be carried through, as there are reasonable but uncontrolled approximations to be made in two steps of the derivation. Nevertheless, the result is quite close to the "ad hoc"ansatz. The calculated OPE energies for 3d and 4f free ions are in qualitative agreement with OPB energies. In Chapter 4, both corrections are implemented in the FPLO scheme to calculate orbital moments in solids. We found that both OPB and OPE corrections implemented in FPLO method, yield reasonably well the orbital magnetic moments of bcc Fe, hcp Co and fcc Ni compared with experiment. In Chapter 5, the effect of spin-orbit coupling and orbital polarization corrections on the spin and orbital magnetism of full-Heusler alloys is investigated by means of local spin density calculations. It is demonstrated, that OP corrections are needed to explain the experimental orbital moments. Model calculations employing one ligand field parameter yield the correct order of magnitude of the orbital moments, but do not account for its quantitative composition dependence. The spin-orbit coupling reduces the degree of spin polarization of the density of states at Fermi level by a few percent. We have shown that the orbital polarization corrections do not change significantly the spin polarization degree at the Fermi level. We also provide arguments that Co2FeSi might not be a half-metal as suggested by recent experiments. In Chapter 6, to understand recent XMCD data for Co impurities in gold, the electronic structure of Co impurities inside gold has been calculated in the framework of local spin density approximation. The orbital and spin magnetic moment have been evaluated. In agreement with experimental findings, the orbital moment is enhanced with respect to Co metal. On the other hand, internal relaxations are found to reduce the orbital moment considerably, whereas the spin moment is less affected. Both OPB and OPE yield a large orbital moment for Co impurities. However, those calculated orbital moments are almost by a factor of two larger than the experimental values. We also found that the orbital magnetic moment of Co may strongly depend on pressure. |
