Understanding and controlling the interfacial magnetic properties of ferromagnetic thin films are crucial for spintronic device applications. However, using conventional magnetometry, it is difficult to detect them separately from the bulk properties. Here, by utilizing tunneling anisotropic magnetoresistance in a single-barrier heterostructure composed of La0.6Sr0.4MnO3 (LSMO)/LaAlO3 (LAO)/Nb-doped SrTiO3 (001), we reveal the presence of a peculiar strong two-fold magnetic anisotropy (MA) along the c direction at the LSMO/LAO interface, which is not observed in bulk LSMO. This MA shows unknown behavior that the easy magnetization axis rotates by 90° at an energy of 0.2 eV below the Fermi level in LSMO. We attribute this phenomenon to the transition between the eg and t2g bands at the LSMO interface. Our finding and approach to understanding the energy dependence of the MA demonstrate a new possibility of efficient control of the interfacial magnetic properties by controlling the band structures of oxide heterostructures.
Control of magnetic anisotropy (MA) is crucial for low-power magnetization reversal in magnetic thin films. From the perspectives of energy efficiency and scalability, gate-voltage control of the MA via modulation of the carrier density and thus, the Fermi level, is highly desirable1,2,3,4. For efficient control of MA and for developing materials that are suitable for the MA control, it is necessary to understand the MA of the magnetic thin films over a wide energy range; however, there are few studies from this point of view. In ferromagnetic (FM) materials, the MA energy is related to the magnetization-direction dependence of the density of states (DOS) via the spin orbit interaction5. Tunneling anisotropic magnetoresistance (TAMR) is a phenomenon observed in tunnel diodes composed of ferromagnetic (FM) layer/tunnel barrier/nonmagnetic (NM) electrode. TAMR is defined as the change of the tunnel resistance or conductance dI/dV, which is proportional to the DOS of the electrodes, when rotating the magnetization of the FM layer5,6,7,8,9,10. Thus, TAMR is useful to understand the magnetic-field direction dependence of the DOS. By measuring TAMR at various bias voltages, one can obtain a high-resolution carrier-energy-resolved map of MA of the FM layer8,9,10.
An equally important aspect of TAMR is that it reflects the DOS at the tunneling interface of the FM layer, and thus it provides a sensitive probe of the interfacial magnetic properties. Thin film interfaces present both problems and opportunities for exploring new functional devices. As a good example, the “dead layer” at the interface of the perovskite oxide La1−x Sr x MnO3 (LSMO, the Sr content x = 0.3–0.4), which is one of the most promising materials due to its intriguing magnetic and electrical properties such as the colossal magnetoresistance11, 12, half-metallic band structure13, 14, and high Curie temperature (TC ~ 370 K)11, is a serious problem for its device applications. For the formation of the dead layer, various possible origins have been proposed, such as intermixing of atoms15, 16, oxygen vacancies17, lattice distortion18,19,20, and MnO6 oxygen octahedral rotations (OOR)21,22,23, which induce orbital, charge, and spin reconstruction at the interfaces of LSMO. These studies on dead layers, however, suggest new ways for controlling the interfacial properties at an atomic level, which are not available in the bulk. To this end, the characterization of the interfacial magnetic properties is highly demanded, but it is difficult with conventional magnetometry because the interfacial properties are usually concealed by the dominant signals from the bulk. Here, by utilizing TAMR in an LSMO/LAO/Nb:STO junction, we obtain the carrier-energy dependence of MA of LSMO for the first time. We also reveal a peculiar strong two-fold symmetry component of MA at the LSMO/LAO interface, which is not observed in bulk LSMO. Moreover, this interfacial MA shows unknown behavior that the symmetry axis of this interface MA rotates by 90° at an energy of 0.2 eV below the Fermi level in LSMO. We attribute this phenomenon to the transition between the eg and t2g bands at the LSMO interface. Our results suggest that controlling the band structure at interfaces will pave a new way for efficient control of the magnetization of FM thin films, which is essential for devices with low-power consumption.
Sample preparation and characterizations
The heterostructure used in this study consists of LSMO (40 unit cell (u.c.) = 15.6 nm)/LaAlO3 (LAO, 4 u.c. = 1.6 nm) grown on a TiO2-terminated Nb-doped SrTiO3 (001) substrate (Nb:STO, Nb 0.5% wt.) by molecular beam epitaxy (MBE) (see Fig. 1a and Methods)24, 25. The in-situ reflection high-energy electron diffraction (RHEED) patterns in the  direction of the 4-u.c. LAO and 40-u.c. LSMO layers show streaky patterns, and especially LSMO exhibits a bright pattern (Fig. 1b), indicating that the sample surface is atomically flat. In fact, the atomic force microscopy measurements show flat terraces and atomic steps with a height of ~0.4 nm, which is equal to one pseudocubic u.c. (Fig. 1c). In the x-ray reciprocal lattice map of the sample measured around the (204)c and reflections of the Nb:STO substrate at room temperature, we see two weaker peaks corresponding to the (260)o and (620)o reflections of the LSMO epilayer (we use the subscripts c and o for the pseudocubic and the orthorhombic crystal structures, respectively) (Fig. 1d). These results confirm that the LSMO layer is coherently grown with respect to the Nb:STO substrate. The (260)o and (620)o peaks of LSMO have nearly the same out-of-plane reciprocal lattice vector Q⊥, indicating that the (260)o and (620)o atomic plane spacings are equal. This is consistent with the common reports on LSMO thin films grown under tensile strain, indicating that the strain effect in LSMO is accommodated equally between the c and c directions22.
For tunneling transport measurements, 600 × 700 μm2 mesas were formed by standard photolithography and Ar ion milling. The bias polarity is defined so that the current flows from the LSMO layer to the Nb:STO substrate in the positive bias.
Magnetic anisotropy components of the LSMO/LAO interface
Figure 2a shows the conduction band (CB) profiles of the LSMO/LAO/Nb:STO tunnel diode under positive and negative bias voltages V. The Fermi level EF is located at 10–20 meV above the CB bottom of STO due to the Nb doping (0.5% wt., the electron density n = 1 × 1020 cm−3)26, while EF lies in the CB formed by the Mn 3d-eg states in LSMO. The LAO layer serves as a tunnel barrier with a height of ~2.4 eV for STO and ~2 eV for LSMO27. TAMR measurements were conducted as follows: dI/dV − V curves were measured at 4 K while applying a strong external magnetic field of 1 T, which aligned the magnetization direction parallel to the magnetic field, in various in-plane directions with an angle step of 10°. The change in dI/dV when rotating the external magnetic field is attributed to the change in the DOS at the LSMO/LAO interface or the LAO/STO interface. As illustrated in Fig. 2a, at positive (negative) V, electrons tunnel from Nb:STO to LSMO (from LSMO to Nb:STO), and thus dI/dV probes the DOS of unoccupied (occupied) states in LSMO. At each V and Φ, where Φ is the angle of the magnetization direction from the c axis in the counter-clockwise direction in the film plane, we define as (%), and is defined as averaged over Φ at V. As seen in the polar plots of as a function of Φ in Fig. 2b, the magnetization direction dependence of the DOS has a mainly two-fold symmetry, but the symmetry axes are different depending on V: At V = −0.1 V (left panel), the maximum is located at ~150°, between the c and axes, while it is at 45° (the c axis) when V = −0.35 V (right panel). The whole picture of this behavior for all V is represented in Fig. 2c, where Δ(dI/dV) is plotted as a function of Φ at V ranging from −0.5 to 0.5 V. This plot shows a peculiar behavior that the symmetry axis changes at V ~ −0.2 V. We fit the data at each bias V using the following equation:
Here C4〈110〉 is the four-fold component with the symmetry axis of 〈110〉c, C2 and C2 are the two-fold components with the c and c axes, respectively. As shown in Fig. 2b, the fitting curves (red curves) well reproduce the experimental data (blue points). The three anisotropy components estimated in the whole range of V are summarized in Fig. 2d. The two-fold symmetry (C2 or C2) is stronger than the four-fold symmetry (C4〈110〉]) in almost all the bias region. The C4〈110〉 and C2 components show a similar V-dependence, stretching from −0.45 V to 0.4 V. The sign of C2 changes at V ~ −0.2 V, which indicates an opposite dependence of the DOS on the magnetization direction between the regions of V > −0.2 V and V < −0.2 V. Because the change in the DOS when changing the magnetization direction, i.e. Δ(dI/dV), is proportional to the MA energy5, the sign change of C2 indicates a 90°-shift of the magnetization direction where the MA energy becomes minimum, and consequently indicates a 90°-shift of the easy magnetization axis of the C2 component from the c to the directions at V ~ −0.2 V when decreasing V. This is the main origin of the symmetry axis rotation observed in Fig. 2b and c.
We discuss the origins of these anisotropy components. The biaxial (four-fold) MA along the 〈110〉c axes and the uniaxial (two-fold) MA along the c axis have been reported for LSMO films grown on STO (001) substrates23, 28,29,30, and thus they are bulk-like properties. The biaxial component originates from the in-plane cubic symmetry of LSMO thin films grown on STO. The uniaxial MA along the c axis has been attributed to various origins such as step edges28, 29 or different OOR between the c and c directions23, 30. In our study, this uniaxial MA along the c axis may also be partially contributed from the high-mobility two dimensional electron gas possibly formed at the LAO/STO interface, where the magnetoconductance has been reported to possess the same two-fold symmetry31. On the other hand, the uniaxial MA component along the c direction has never been reported for LSMO. Because the crystal structure of LSMO grown on a substrate with a cubic symmetry such as STO is equivalent between the c and directions, the uniaxial MA along the c direction is not a bulk property, and thus must be attributed to the LSMO/LAO interface.
To confirm the interface origin of the uniaxial MA component along the c axis, we measured the planar Hall resistance (PHR) of a Hall bar with a size of 50 × 200 μm2 formed along the c direction of a reference sample composed of LSMO (40 u.c.)/LAO (4 u.c.). This sample was grown on a non-doped STO (001) substrate under the same conditions as those for the sample used for the TAMR measurements. PHR is proportional to sin Φ cos Φ, where Φ is the angle between the magnetization and the current direction32. In Fig. 3, we show the PHR measured for the reference sample at various magnetic field directions, where θ is the angle between the magnetic field and the current flown in the c axis. ΔR is defined as the PHR with respect to the one at a zero magnetic field. In contrast to the TAMR results, the PHRs of the reference sample measured when the magnetic field is parallel to the c and directions are identical (i.e. four-fold like) except the opposite signs, showing no clue of the uniaxial MA along the c axis. On the other hand, the PHRs clearly show a dominant uniaxial MA with the easy axis along the c direction. Because PHR mainly reflects the bulk properties of LSMO, our results confirm that the C4〈110〉 and C2 components are the MA inherited from bulk LSMO, while the C2 component is the MA that appears only at the LSMO/LAO interface.
Because the C2 component is attributed to a symmetry breaking between the c and directions occurring locally near the LSMO/LAO interface, the OOR mechanism is most likely the origin of C2. Recently, it has been clarified that adjacent corner-sharing oxygen octahedra in LAO grown on STO (001) rotate in the opposite directions around the axis33, 34, and that the OOR in the underlayer is transferred to the first 3–4 u.c. layers of LSMO21, 22 (Fig. 4a). Figure 4b shows four adjacent MnO6 octahedra in a (001)c plane of LSMO near the LSMO/LAO interface. We see that under the OOR around the axis illustrated in Fig. 4a, the oxygen octahedra located along the c direction rotate in the same direction (see the green oxygen spheres around Mn1 and Mn3). This rotation direction is the opposite to that of the adjacent rows of Mn atoms (see Mn2 and Mn4). When Fig. 4b is projected in the plane as shown in Fig. 4c, one can see that the vertical O–Mn1–O and O–Mn3–O bonds (see the green spheres) remain nearly perpendicular to the (001)c plane (dotted line). Because the vertical O–Mn1–O bonds are rotated in the same direction as the O–Mn3–O bonds, the hopping integral (t13) between Mn1 and Mn3 remains nearly unchanged from the value of the bulk. On the other hand, when Fig. 4b is projected in the (110)c plane (Fig. 4d), one can see that the MnO6 oxygen octahedra around Mn2 and Mn4 are largely tilted to the left. This decreases the hopping integral (t24) between Mn2 and Mn4. The difference between t13 and t24 yields the anisotropic DOS between the c and directions and consequently induces C2 at the LSMO/LAO interface.
The most striking feature found in our study is the sign reversal of C2 at V = −0.2 V (Fig. 2d). This behavior is likely related to the band structure of LSMO, as explained below. In Fig. 2d, one can see that C4〈110〉 and C2 show similar V dependence in all the V region. These results indicate that both originate from the same band located around EF of LSMO, i.e. the up-spin Mn 3d-eg band. The C4〈110〉 and C2 components disappear at ~V = −0.45 V, which means that EF is located at ~0.45 eV above the bottom of the eg band. This is consistent with the results of angle-resolved photoemission spectroscopy (ARPES) measurements for LSMO35. Therefore, the emergence of positive C2 below V = −0.2 V is likely associated with the t2g band, which is located below the eg band. Although the t2g state is located at 0.5–1 eV below EF in bulk LSMO35, it is thought to be largely pushed up by the polar mismatch at the LSMO/LAO interface36. Thus, we attribute the sign change of C2 to the transition from the eg band (V > −0.2 V) to the t2g band (V < −0.2 V) at the LSMO interface. As mentioned above, due to the OOR at the LSMO/LAO interface, the DOSs of both the eg and t2g bands in the c direction are larger than those in the direction. However, the relationship between the DOS and the magnetization in these two band components is opposite: It is known that the electron transfer via the eg orbitals enhances the double exchange interaction and strengthens the ferromagnetism, while the one between the t2g orbitals enhances the super-exchange interaction and weakens the ferromagnetism37. Therefore, the enhancement of DOS in the c direction relative to that in the direction makes the c axis the easy magnetization direction in the case of the eg orbitals, while hard magnetization direction in the case of the t2g orbitals. The transition between these two bands at V = −0.2 V thus leads to the opposite magnetization direction-dependence of the DOS, as consequently observed by TAMR5.
In summary, using TAMR measurements, we have successfully obtained a high-resolution map of the MA spectrum of LSMO for the first time. In addition to the biaxial MA along 〈100〉c and the uniaxial MA along c, which originate from bulk LSMO, we found a peculiar uniaxial MA along the c, which is attributed to the LSMO/LAO interface. The symmetry axis of this interface MA rotates by 90° at an energy of 0.2 eV below EF of LSMO, which is attributed to the transition from the eg band (>−0.2 eV) to the t2g band (<−0.2 eV). These findings hint an efficient way to control the magnetization at the LSMO thin film interfaces, as well as confirm the rich of hidden properties at thin film interfaces that can be revealed only by interface-sensitive probes. This work also suggests the use of the TAMR measurement as a simple but highly sensitive method for characterizing interfacial magnetic properties of magnetic tunnel junctions, which is important for developing spintronic devices.
The heterostructure used in this study consists of LSMO (40 unit cell (u.c.) = 15.6 nm)/LaAlO3 (LAO, 4 u.c. = 1.6 nm) grown on a TiO2-terminated Nb-doped SrTiO3 (001) substrate (Nb:STO, Nb 0.5% wt.) by molecular beam epitaxy (MBE) with a shuttered growth technique24, 25. The fluxes of La, Sr, Mn, and Al were supplied by Knudsen cells. The LAO and LSMO layers were grown at 730 °C with a background pressure of 2 × 10−4 Pa of a mixture of oxygen (80%) and ozone (20%). After the growth, the sample was further annealed at 600 °C in ambient atmosphere for 1 hour to reduce the density of oxygen vacancies.
For tunneling transport measurements, a 50-nm-thick Au film was deposited on top of the sample, and 600 × 700 μm2 mesas were then formed by standard photolithography and Ar ion milling. Au wires were bonded to the Au electrode and the backside of the Nb:STO substrate by indium.
The dI/dV-V characteristics were numerically obtained from the I-V data with a differential interval of 10 mV. See Supplementary Information for more details on how to extract the data plotted in Fig. 2.
The datasets of the current study are available from the corresponding author on reasonable request.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This work was partly supported by Grants-in-Aid for Scientific Research (No. 26249039, No. 23000010) and Project for Developing Innovation Systems of MEXT, Spintronics Research Network of Japan (Spin-RNJ), and the Cooperative Research Project Program of RIEC, Tohoku University. We thank K. Takeshima and T. Matou for technical helps in sample growth.
Electronic supplementary material
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.