Uniaxial Magnetocrystalline Anisotropy of Metal/Semiconductor Interfaces: Fe=ZnSe001 Elisabeth Sjo ¨stedt, Lars Nordstro ¨m, Fredrik Gustavsson, and Olle Eriksson Department of Physics, Uppsala University, Box 530, 751 21 Uppsala, Sweden (Received 4 June 2002; published 10 December 2002) A theoretical study of the magnetic moments and the in-plane magnetic anisotropy of an interface between a cubic ferromagnet and a cubic semiconductor, Fe=ZnSe001, is presented. Theory confirms the observed, much debated, uniaxial anisotropy of the iron film. This result is important since the calculations are for perfect interfaces with squarelike environments, proving that the fourfolded symmetry of the interface Fe atoms is broken beyond the nearest neighboring semiconducting layer, effects that are usually assumed small. It is demonstrated how the uniaxial anisotropy is produced by the directional covalent bonds at the interface, even without atomic relaxations. DOI: 10.1103/PhysRevLett.89.267203 PACS numbers: 75.70.–i, 75.10.Lp, 75.30.Gw Heterostructures consisting of ferromagnetic materials layered with semiconducting materials (FM/SC) show a range of new and interesting features, with potential use in spin electronics [1]. It has been suggested that an appropriate utilization of such materials will result in novel technologies based on spin-FET (field effect tran- sistor), spin-LED (light emitting diode), etc. In addition, such heterostructures are expected to stand as strong can- didates in the search for the next generation of giant and colossal magnetoresistance materials, where the transport properties are heavily dependent on an applied field [2]. Recently, Fe films grown on semiconductors in the cubic zinc-blende structure, e.g., ZnSe, GaAs, or InAs, have been in focus as potentially useful material combi- nations [3–7]. One of the key scientific issues has been identified with the atomistic/microscopic properties of the interface of these materials. In contrast to the ex- pected fourfold local symmetry of the interface, an in- plane uniaxial magnetic anisotropy (UMA) was observed for thin Fe layers, with easy axis along [110] or 1 10, depending on the specific system [3–7]. As the thickness of the Fe layer increases, the symmetry changes gradually into the fourfold symmetry of bulk bcc Fe, with equiva- lent easy axes along the h100i azimuths. The effect has been found for various semiconductor substrates, and, in particular, for Fe=GaAs and Fe=ZnSe. There have been many theories attempting to explain the microscopic origin of this UMA. Some of them, such as those based on the formation of an interface alloy or a large density of interface steps, have been ruled out. Today, the most favored explanation involves the recon- struction of the SC surface, due to the dimerization of its dangling bonds. These reconstructions lead to atomic rows along either [110] or 1 10, depending on the termi- nation. The argument is that these reconstructions of the substrate remain after the growth of the FM film, which would produce a large uniaxial strain in the magnetic film. This strain in turn causes a uniaxial anisotropy through the magnetoelastic effect. Alternatively, the ori- gin of the UMA has been attributed to the anisotropic interface bonding. For instance, by means of electronic structure calculations it was found that the two Fe atoms at the interface, although showing very similar densities of states, were affected by their inequivalent positions relative to the sp 3 bonds of the semiconductor [8]. In recent years, there have been experimental efforts to determine which of the two models causes the in-plane UMA. On one hand, Kneedler et al. [4] showed that the easy axis is independent of the direction of the sur- face reconstructions for two different Fe=GaAs systems, indicating that the strain in the Fe film has little effect. On the other hand, Xu et al. [5] deduced that it is the difference in the magnetoelasticity that causes the differ- ence in easy axes observed for Fe=GaAs and Fe=InAs. Neither of these studies can completely rule out the competing model. However, if an interface with square geometry would show a UMA, the effect of any strain would only marginally affect its magnitude relative to the unstrained film. This is in contrast to the situation in, e.g., bcc Fe, where a symmetry breaking strain is crucial for the UMA. It is the purpose of the present Letter, to show by means of first principles calculations, that ideal unstrained magnetic films in a FM/SC system do exhibit large in- plane UMA. As a model system we have chosen Fe=ZnSe, since it has favorable qualities which simplifies comparisons with experiments; i.e., the system has a good lattice matching and, due to the lower reactivity of ZnSe as compared to, e.g., GaAs, has experimentally very clean interfaces. In addition, the surface reconstructions of this II-VI SC are comparatively simple. The Se-terminated surface has full coverage and is 2  1 reconstructed due to dimeriza- tion, while the Zn terminated surface only has half cover- age leading to a c2  2 structure [9]. Prior to this work there have been several theoretical studies of the Fe=ZnSe system. Continenza et al. per- formed semirelativistic spin-polarized calculations for three different Fe=ZnSe supercells, repeated to form multilayered structures [8]. The electronic structure and VOLUME 89, NUMBER 26 PHYSICAL REVIEW LETTERS 23 DECEMBER 2002 267203-1 0031-9007= 02=89(26)=267203(4)$20.00  2002 The American Physical Society 267203-1