Abstract : Magnetite Fe3O4 is a promising material for spintronics since band structure calculations predict it to be half-metallic. One thus expect large magnetoresistance in magnetic tunnel junctions using Fe3O4 as electrode. We have grown 3 to 50 nm-thick Fe3O4 thin films onto sapphire by molecular beam epitaxy. The films are single crystalline but comprise a large number of antiphase boundaries. These defects are the origin of the magnetic anomalies of Fe3O4 thin films. which we qualitatively reproduiced
by using a one-dimensionnal model, the results being compared with the characteristic size of the antiphase boundaries measured by a fractal analysis. Moreover, the Verwey transition is not seen for thicknesses below 20 nm because of finite size effects; all samples show slow magnetization dynamics.
We have also developped a growth method allowing us to deposit γ-Al2O3 epitaxially onto Fe3O4 while controling the stoichiometry at the interface. Al2O3 thin films of thicknesses greater than 2 nm are continuous and allow magnetic decoupling between electrodes. Direct measurements by spin resolved photoemission yields to a -40 % the spin polarization at the Fe3O4/γ-Al2O3 interface whereas the tunnel magnetoresistance (reduiced by the magnetic disorder generated by the antiphase boundaries) is +3 % at 300 K for Fe3O4/γ-Al2O3/Co magnetic tunnel junctions.