Growth and physical properties of Fe3O4(111)-based epitaxial tunnel junctions

Figure 1: HRTEM micrograph of the Fe3O4 (15 nm)/γ-Al2O3 (1.5 nm) bilayer. The similarity of the two lattices (spinel) demonstrates that the Al2O3 layer grows in the γ phase (Coll. P. Warin, P. Bayle-Guillemaud DRFMC-SP2M). Inset displays the RHEED pattern along the [1-100] direction of the crystalline γ-Al2O3 layer at the end of the growth.

Mastering the growth of epitaxial Fe₃O₄(111) thin films allows us to integrate this oxide as electrode in a magnetic tunnel junction (MTJ). The literature reports a number of efforts to elaborate MTJs using a magnetite electrode. However up to now, the best results reported have been obtained in MTJs containing an AlO_x alumina amorphous barrier. As a matter of fact, the tunnel magnetoresistance (TMR) effects are mainly described by the electrons tunnelling from the interfaces and therefore strongly depend on the quality of the interfaces. We have chosen to study the epitaxial growth of a MTJ with a crystalline γ-Al₂O₃(111) barrier in order to have a better understanding of the spin-dependent tunnelling transport. An important part of this work consisted to analyze the structural, chemical and physical properties at the Fe₃O₄(111)/γ-Al₂O₃(111) interface [1]. The spin polarization at this interface has also been studied by two different techniques: spin-resolved photoemission and TMR measurements.

At first we have studied the growth of fully epitaxial Fe₃O₄(111)/γ-Al₂O₃(111) bilayers to be included in a MTJ. Epitaxial growth by molecular beam epitaxy (MBE) of a γ-Al₂O₃ ultra-thin layer (a few nanometers) on top of a magnetite electrode was achieved by evaporation of metallic aluminium from a Knudsen effusion cell under atomic oxygen. The RHEED diffraction patterns (inset of figure 1) recording during the growth demonstrated the two-dimensional growth mode of Al₂O₃. In situ Auger electron spectroscopy (AES) and X-ray photoelectron spectroscopy (XPS) measurements (Fe 2p and Al 2p core levels) have also been performed in order to check the oxidation state of Fe at the Fe₃O₄/Al₂O₃ interface and the full homogeneous oxidation of aluminium. In appropriate growth conditions (Al flux, atomic oxygen pressure), we were able to obtain a stoichiometric Fe₃O₄/Al₂O₃ interface [1,2]. The magnetic properties of these stoichiometric bilayers, obtained by VSM magnetometry and XMCD analyses (not shown here), exhibited the same magnetic behavior (remanence, coercivity, Verwey transition, magnetic moment) as the corresponding uncovered Fe₃O₄ layers. The HRTEM cross section in figure 1 confirmed the very good epitaxy of the Fe₃O₄(111)/γ-Al₂O₃(111) system. Finally, the insulating properties of the γ-Al₂O₃(111) ultra-thin layers have been measured using a conductive tip AFM (Coll. UMR CNRS/Thales) validating the tunnel barriers properties [1,2]. No pinholes have been detected, demonstrating that the Fe₃O₄ layer is entirely covered by a 1.5 nm thick Al₂O₃ layer.

In a second step, we have studied the spin polarization (P) of the Fe₃O₄(111)/γ-Al₂O₃(111) system defined by the following relationship: P= (N_up(E_F)-N_down(E_F))/(N_up(E_F)+N_down(E_F))

N_up(down) (E_F) corresponding to the electron density of states at the Fermi level for the spin up, (down) direction.

Despite different experimental works, there is not yet a consensus on the value of the spin polarization obtained on Fe₃O₄ surfaces. Spin resolved photoemission studies [3] reported values from -80 % to +16 %. Furthermore, the pertinent parameter for spin dependent tunnelling is not the spin polarization of Fe₃O₄ free surfaces but that of the Fe₃O₄/tunnel barrier interface. Several analyses have shown that the value and sign of the spin polarization strongly depends on the nature of the tunnel barrier. This is why we have performed spin resolved photoemission on a Fe₃O₄ (25 nm)/γ-Al₂O₃ (2 nm) bilayer. The spin polarized photoemission results, obtained at the ESRF synchrotron (ID 08 beamline), are displayed in Figure 2. The spin polarization at the Fe₃O₄/γ-Al₂O₃ is negative, about -20 %. However these experiments have been performed in zero magnetic field. Taking into account the remanence correction (M_r/M(6 T)~ 50 % in our 25 nm thick magnetite film), we have obtained a value of -40% for the spin polarization of the Fe₃O₄/γ-Al₂O₃ interface [4]. This value is lower than predicted by band structure calculations for Fe₃O₄ but different explanations can be put forward (see references 1, 4). The low energy resolution of the experiment (0.7 eV), in comparison with the spin gap of magnetite, or the contamination of the γ-Al₂O₃ layer (no cleaning has been performed for this measurement) which could give rise to an unpolarized signal, can lower the value of the measured spin polarization. Nonetheless, the negative sign of the spin polarization is coherent with band structure calculations and shows that the γ-Al₂O₃ barrier does not change the sign of the spin polarization of Fe₃O₄.

Figure 2: Spin resolved photoemission spectra of the valence band of the Fe3O4 (25 nm)/γ-Al2O3 (2 nm) bilayer. Upper panel: spin up and spin down photoemission intensities. Lower panel: spin polarization (Coll. ESRF-ID08).

Figure 3: Room temperature magnetoresistance (TMR) of a 96 μm2 Fe3O4 (25 nm)/γ-Al2O3 (2 nm)/Co (15 nm) magnetic tunnel junction (Coll. R. Mattana, P. Seneor, K. Bouzehouane, F. Petroff, UMR CNRS/Thalès).

Finally, MTJs were prepared by covering the Fe₃O₄/γ-Al₂O₃ bilayers with an iron or cobalt counter-electrode. The magnetic hysteresis loop of a Fe₃O₄(25 nm)/γ-Al₂O₃(3 nm)/Co(15 nm) trilayer shows a clear magnetic decoupling between the two ferromagnetic electrodes. Figure 3 displays a TMR curve obtained on a 96 mm² MTJ (coll. UMR CNRS/Thales) at room temperature. The first abrupt switching around 50 Oe corresponds to the cobalt layer and the second switching around 600 Oe is related to Fe₃O₄ with probably a slight magnetic coupling. The first TMR measured, normalized with respect to R(μ₀H= 0), is around +3 % at room temperature [1, 5] and does not depend on temperature in the range 130 K–300 K. Considering the Jullière’s model, this positive TMR seems to be in contradiction with the negative spin polarization of the Fe₃O₄/γ-Al₂O₃ determined by spin polarized photoemission and the positive spin polarization of the Co/Al₂O₃ interface generally observed. Different studies are in progress in order to understand these first results (effects of APBs, lithography, …) and to optimize the absolute value of the TMR.

REFERENCES:

[1] A. M. Bataille, Ph.D. thesis, University Paris XI (2005).

[2] A. M. Bataille et al., Appl. Phys. Lett. 86, 1 (2005).

[3] H.-J. Kim et al., Phys. Rev. B 61, 15288 (2000). Y. S. Dedkov et al., Phys. Rev. B 65, 064417 (2002).

[4] A. M. Bataille et al., Phys. Rev. B 73, 172201 (2006). [5] A. M. Bataille et al., J. Magn. Magn. Mater. (2006).

#788 - Last update : 10/13 2009