Spin-resolved photoelectron spectroscopy of Fe3O4---revisited

Recently Tobin et al (2007 J. Phys.: Condens. Matter 19 315218) reported on the spin-resolved photoemission study of Fe3O4(001) films, claiming magnetite being a case against half-metallicity. In the present communication we re-examine recent spin-resolved photoemission experiments on Fe3O4 and explain why their criticism is unfounded. The materials class of half-metallic ferromagnets comprising Heusler alloys [1, 2] and transition metal oxides such as CrO2 [3] and Fe3O4 [4, 5] attracts continued interest because of its intriguing potential as spin-injector material in magnetic tunnel junctions or in spintronic devices. A particular challenging case is Fe3O4, which can be grown epitaxially as thin films with (111) and (001) orientation. The spin polarization values of both orientations at the Fermi energy determined by spin-polarized photoelectron spectroscopy (SPPES) have given rise to different conclusions. Interpretations were put forward in the framework of either the ionicconfiguration-based approach [5] or a band-type approach [4]. Such data have also been interpreted as proof against half-metallicity of Fe3O4 [5]. However, STM and electronic structure calculations provided evidence of surface reconstructions, at least for the (001) orientation [6], pointing to the limited applicability of surface sensitive PES measurements at low photon energies to elucidate bulk properties. Our SP-PES investigations of Fe3O4(111)/W(110) showed at room temperature a spin polarization of −80% near the Fermi energy, EF [7] (see figure 1). In this paper we pointed out (p 64417-3, second-to-last paragraph) that a ‘spin polarization value by itself is no proof of a half-metallic state’. This caution on our part was obvious in view of the reference material Fe(110) showing also −80% spin polarization near EF. In contrast, Tobin et al [5] claimed that we had concluded from the small region in k-space accessible by our experiment the half-metallic nature of Fe3O4. This is definitely incorrect. Instead, we took the result of P = −80% as indicative to abolish the ionic-configuration-based approach [8], setting an upper limit of P = −66.6%, and followed a band-type description. Indeed, distinct dispersion was found for O(p)-derived and Fe(d)-derived bands of Fe3O4(111) [9]. By correlating spinresolved features in the photoemission spectra with those in the electronic band structure calculations, very good agreement was found along the [111] direction [7]. We took these facts and the observed emergence of a band gap in the spin-up spectra upon oxidation of the Fe(110) film to yield Fe3O4(111) [7] as evidence for a half-metallic state of Fe3O4 with a (111) orientation. The (001) orientation of Fe3O4 is yet another interesting case in the present context. It has been pointed out in the literature that strain in Fe3O4 films may play a crucial role and affect the electron spectroscopy data. A good measure of strain relief in thin films is the sharpness of the temperaturedependent Verwey transition near 120 K. As an example we show in figure 2 the magnetization as function of temperature of one of our epitaxial Fe3O4(001) films grown on MgO(001), confirming the very high quality of the films. In contrast, the Verwey transition of a so-called strain relieved Fe3O4 film in figure 2 of [5] is very much smeared out.