Spin reorientation at (110)-La 2/3 Sr 1/3 MnO 3 / LaCoO 3 interfaces by orbital/charge reconstruction

The interface reconstruction in perovskite heterostructures caused by interfacial octahedral tilt/rotation and its effects on the spin, charge, and orbital degrees of freedom is a very attractive topic for correlated oxides. Here, we present a systematic investigation on tensely strained (110)-LaCoO 3 /La 2/3 Sr 1/3 MnO 3 /LaCoO 3 trilayers, focusing on orbital reconstruction and accompanied effects. The most remarkable finding is the reordering of the energy levels of Mn-3 d orbitals at the interface: the low-lying orbital becomes d x 2 -y 2 for sandwiched La 2/3 Sr 1/3 MnO 3 rather than d 3z 2 -r 2 as expected for a bare La 2/3 Sr 1/3 MnO 3 film. Interlayer charge transfer via d x 2 -y 2 orbitals is further detected as a driving force of orbital reconstruction. Due to spin–orbit coupling, the charge/orbital reconstruction produces a chain effect on the spin degree of freedom of the La 2/3 Sr 1/3 MnO 3 layer, resulting in a dramatic spin reorientation by 90 ○ in a film plane. The present work demonstrates how to tune macroscopic properties of correlated oxides via mutual coupling between different degrees of freedom.


INTRODUCTION
Perovskite transition metal oxides (TMOs) with strongly coupled spin, charge, and orbital degrees of freedom provide a valuable playground for the exploration for emergent phenomena. [1][2][3][4] What is of special importance is that the heterostructures are composed of different TMOs, which own unique interface phases with cooperatively distorted oxygen octahedra and reconstructed spin/charge/orbital orders. [4][5][6][7][8][9][10] There are reports on dramatic variations in macroscopic properties for (001) oriented multilayers, produced by even a subtle octahedral tilt/rotation at interfaces. [11][12][13][14] As shown by Liao et al., 15 through transferring the octahedral rotation in NdGaO 3 to the La 2/3 Sr 1/3 MnO 3 (LSMO) film, a giant anisotropic transport in ultrathin LSMO films as well as a realignment of the magnetic easy axis can result. In a recent work, Zhang et al. 16 demonstrated how symmetry mismatch drove a spin reorientation for the LSMO/LaCoO 2.5 heterostructures. It was found that, at the interface, MnO 6 octahedra share the apical oxygen with neighboring CoO 4 tetrahedra, forming elongated octahedra which support perpendicular magnetic anisotropy. The strong effect of the engineered interface was also observed in magnetic oxides other than LSMO. As reported by Kan et al., 17 the SrRuO 3 /Ca 0.5 Sr 0.5 TiO 3 combination led to a large Ru-O-Ti bond angle; thus, a unique SrRuO 3 phase with a substantially larger Ru-O-Ru bond angle than that of the bulk counterpart is observed.
These works clearly demonstrate the effects of octahedral tilt/rotation on spin ordering. Different from its bulk counterpart, the interface phase suffers from a spatial confinement in the outof-plane direction. In addition to enhancing quantum fluctuation, this feature will allow a full use of the advantage of interlayer engineering, getting states available for either constituent of the heterostructure. [18][19][20] In general, interface engineering takes the effect via modifying the multiple degrees of freedom of the interface phase. 4,[21][22][23] Due to the strong coupling between different degrees of freedom, any variation in one degree of freedom will cause a chain effect. Undoubtedly, a deep understanding of the interface effect will strengthen our capabilities to design materials on demand. In this work, we presented a systematic investigation on tensely strained (110)-LCO/LSMO/LCO trilayers (LCO = LaCoO 3 ), focusing on interface orbital reconstruction and accompanied effects associated with spin degrees of freedom. As reported by Chen et al., the (110)-orientated films exhibit a faster strain relaxation process along the [110] direction than that along the [001] direction. 24 That is to say that the tensile strain along the [001] direction dominates the lattice distortion of the LSMO film deposited on the (110)-STO substrate. This causes the elongation of MnO 6 octahedra along the in-plane [001] axis and the preferred occupation of the d 3z 2 -r 2 orbital. 25,26 Thus, the magnetic easy-axis of the (110)-LSMO/STO bare film aligns along the [001] axis. In this case, interlayer orbital hybridization and charge transferring are expected to hardly occur due to the absence of the Mn 3d 3z 2 -r 2 and O 2px (2py) overlap. Surprisingly, we observed a reversion of the energy levels of the d x 2 -y 2 and d 3z 2 -r 2 orbitals: the low-lying orbital is d x 2 -y 2 in LSMO of the trilayers. Accompanying orbital reconstruction, a Mn-to-Co charge transfer via the d x 2 -y 2 orbitals takes place. It is this process that stabilizes the d x 2 -y 2 orbital. Accordingly, the magnetic easy axis of the LSMO layer undergoes a switching from the [001] to the [110] direction due to the strong correlation between the orbital and spin degrees of freedom. Fig. 1(a)] were grown on (110)-SrTiO 3 (STO) single crystal substrates (5 × 3 × 0.5 mm 3 ) using the technique of pulsed laser deposition (PLD). During deposition, the temperature of the substrate was maintained at 700 ○ C (for LSMO) or 635 ○ C (for LCO) and the oxygen pressure was fixed to 30 Pa. Here, a low growth temperature was adopted for the LCO layer to avoid recrystallization, which will cause rough interfaces. The repetition rate of the laser pulse was 2 Hz, and the fluence was 2 J/cm 2 (KrF excimer laser, wavelength = 248 nm). After deposition, the samples were cooled to room temperature at the rate of 10 ○ C/min in an oxygen pressure of 100 Pa. The layer thickness was set to 7 nm for LCO and to 4 nm, 5 nm, 9 nm, 10 nm, 15 nm, and 19 nm for LSMO. For comparison studies, two bare LSMO films with the thicknesses of 6 nm and 10 nm, respectively, were also fabricated. Here, film thickness has been determined by the number of laser pulses, after a  The surface morphology of the trilayers was measured by using an atomic force microscope (AFM, SPI 3800N, Seiko). The crystal structure of the films was determined by using a Bruker x-ray diffractometer equipped with thin film accessories (D8 Discover, Cu Kα radiation). Lattice images were recorded by using a high-resolution scanning transmission electron microscope (STEM) with double C S correctors (JEOL-ARM200F). Magnetic measurements were conducted by using a quantum-designed vibrating sample magnetometer (VSM-SQUID) in the temperature interval from 5 K to 300 K and the magnetic field range up to 7 T.

LCO/LSMO/LCO trilayers [
The XAS spectra were collected at the Beam line BL08U1A in Shanghai Synchrotron Radiation Facility, in the total electron yield mode. The spectra were measured at the Mn L-edge for two polarization directions by setting x-ray polarization to [001] and [110] directions in sequence. The spectra normalization was made by dividing the spectra by a factor such that the L 3 pre-edge and L 2 post-edge have identical intensities for the two polarizations. After that, the pre-edge spectral region was set to zero and the peak at the L 3 edge was set to one. The XLD (I [001] −I [110] ) is the intensity difference of normalized XAS along two measurement directions, which gives information on the empty Mn-3d states. Co L-edge XAS was measured with the x-ray polarization to [001]. The measurement temperature for XAS and XLD is 300 K.  Figure 1(b) shows the θ-2θ x-ray diffraction (XRD) spectra for selected LCO/LSMO/LCO trilayers with the LSMO layer thicknesses of t LSMO = 5 nm, 10 nm, 15 nm, and 19 nm. The XRD spectrum is somewhat complex, composed of multiple broad peaks whose number grows with t LSMO . This is the typical feature of trilayers, arising from the diffraction/interference of xray between three layers. As shown by the red curve in Fig. 1(b), the calculated curve (red line) mimics the experimental one (black line) very well. Multiple XRD peaks are signatures of high crystal quality of the trilayers. According to the results of curve fitting, the out-of-plane lattice parameter defined by d 110 = √ 2/2a 0 can be deduced, where a 0 is the lattice parameter of the perovskite unit cell for LSMO or LCO. It is noted that d 110 ≈ 2.71 Å for LSMO and ∼2.68 Å for LCO, smaller than the bulk value (∼2.74 Å for LSMO and ∼2.70 Å for LCO). As expected, both films are in-plane tensely strained.

RESULTS AND DISCUSSION
To get the information about the in-plane lattice structure, the reciprocal space mappings (RSMs) of the (222) and (130) reflections are measured. As shown in Fig. 1(c), multiple reflections of the trilayers are also detected by RSMs, aligning vertically with that of the substrate. In addition to a fully coherent growth of the film on the substrate, this result indicates that the trilayers share exactly the same in-plane lattice constant with STO, a 001 = 3.905 Å and a 110 = 5.522 Å. Therefore, the trilayers are fully tensely strained. This conclusion is consistent with that drawn from curve fitting of the Θ-2Θ scanning. Similar results are obtained for other trilayers investigated here (not shown). Figure 2(a) shows the typical high-angle annular dark-field (HAADF) image of the cross section of (110)-LCO (7 nm)/LSMO (10 nm)/LCO (7 nm), recorded along the [001] zone axis by using a STEM. Here, the brighter and fainter spots correspond to the La/Sr and Mn/Co atomic columns, respectively. Notably, no structural modulation is observed in the bottom LCO layer neighboring substrate (right side LCO). It means that LCO is close to stoichiometric LaCoO 3 . In contrast, dark stripes appear in the top LCO layer, indicating the formation of the LaCoO 3-δ phase. [27][28][29] From first glance, the interface is clear-cut [ Fig. 2(a) and the left panel of Fig. 2(b)]. A further analysis of electron energy loss spectroscopy (EELS) shows that the interface is really sharp, with only minor interlayer diffusion to the distance within one unit cell [the right panel of Fig. 2(b)].
To get the information on the orbital structure of the Mn-3d electrons, the technique of XAS was adopted.  (110) planes, respectively. The sketch shows the experiment setup. Bottom panels are the corresponding XLD spectra, amplified by a factor of ten. Shaded areas provide the information on orbital occupancy. The preferentially occupied orbital is d 3z 2 -r 2 for the bare film and d x 2 -y 2 for superlattices.
show the Mn-XAS spectra of a bare LSMO film (6 nm in thickness) and a [LCO (3 m)/LSMO (3 m)]5 superlattice, respectively. Here, the superlattice with an ultrathin layer thickness (3 m) was chosen to highlight the interface effect. Since magnetic signals mainly come from the LSMO layer, we only present the Mn-L 2 and L 3 absorption peaks. Two spectra are obtained for each sample by setting x-ray polarization to [001] and [110] directions in sequence. 30 As shown by the upper right sketch in Fig. 3(a), [001] and [110] planes are parallel to the d 3z 2 -r 2 and d x 2 -y 2 orbitals, respectively. 25,26 Two broad peaks are observed in the interested energy range, corresponding to the L 2 and L 3 absorption edges of Mn 3d electrons. 31,32 As reported, the L 2 peak contains important information on the orbital structure: a high (low) peak implies a low (high) orbital occupancy. 8,33 To highlight the difference in the absorption peaks along two directions, x-ray linear dichroism (XLD) spectra, defined by I [001] − I [110] , are calculated, where I [001] and I [110] are the peak intensities along the corresponding directions. As well documented, the integration of the XLD spectrum around the L 2 absorption peak gives a direct measure to the occupancy of the d x 2 -y 2 and d 3z 2 -r 2 orbital states. 34,35 A negative (positive) value means a preferential occupation of the d 3z 2 -r 2 (d x 2 -y 2 ) orbital. For the bare LSMO film, the XLD spectrum exhibits a negative peak around L 2 [the bottom panel of Fig. 3(a)], indicating that the low-lying orbital is d 3z 2 -r 2 when LSMO is in the tensile state. This result is understandable noting that tensile strain will elongate the MnO 6 octahedron along the [001] axis, thus lowering the energy level of d 3z 2 -r 2 . Surprisingly, the orbital occupancy is different in the superlattice: the preferred orbital now is d x 2 -y 2 as implied by the positive XLD peak [ Fig. 3(b)] rather than d 3z 2 -r 2 as required by tensile strains. This result is interesting in a sense that it reveals the occurrence of interface coupling-induced orbital reconstruction.
Orbital reconstruction should mainly take place at the LCO/LSMO interfaces since it is invisible when LSMO is thicker than 10 unit cells (not shown). Obviously, interlayer coupling has introduced a mechanism that overcomes the effect of tensile strains, driving d x 2 -y 2 to a lower energy level than d 3z 2 -r 2 . After a careful analysis of the XAS spectra, we found signatures of Mn-to-Co charge transfer. Compared with a bare LSMO layer, the Mn-L 3 (Co-L 3 ) peak of the superlattices shows a high energy (low energy) shift [ Fig. 4(a)]. This result is indicative of an increase (decrease) in the valence state of Mn ions (Co ions). 36 38 It is this bonding process that lowers the energy level of d x 2 -y 2 [ Fig. 4(b)]. Obviously, charge and orbital reconstructions concomitantly take place at interfaces 7-9 because of the strong coupling between the corresponding degrees of freedom. As for the d 3z 2 -r 2 orbital, it lies in the film plane and cannot form a chemical bond with the d x 2 -y 2 or d 3z 2 -r 2 orbital of the neighboring lattice plane. Therefore, it has no contributions to interlayer coupling. From this figure, we can understand the orbital reconstruction in the superlattice. In fact, we have estimated the energy gained by chemical bonding and found that it is ∼1.5 eV, 40 i.e., the d x 2 -y 2 will be considerably lowered by the formation of a covalent bond.
As demonstrated above, by grouping LSMO together with LCO, the orbital/charge structure at the interface is amended. An interesting issue is that how this interface reconstruction affects the spin degree of freedom. Fascinatingly, we found a concomitant spin reorientation: after the orbital/charge reconstruction, spin orientation rotates by 90 ○ in the film plane from the [001] to [110] axis. For a bare LSMO film, as shown in Fig. 5 Similar to trilayers, the easy axis of the superlattices is also along the [110] direction with an even higher anisotropy constant. This is reasonable since superlattices own more interfaces than trilayers (see Fig. S3 of the supplementary material for magnetic data of the superlattice).
A further issue to be addressed is that how the orbital reconstruction modifies the spin orientation. As mentioned above, LSMO owns strongly coupled spin, charge, and orbital degrees of freedom. As well documented, the anisotropy of the spinorbit energy is directly related to the anisotropy of the orbital moment, 41 ). Accordingly, the easy and hard axes are reversed (see Note 1 of the supplementary material for detailed calculations).  To quantitatively describe magnetic anisotropy, we converted the M-T curves in Fig. 5  direction, the saturation state is not gained until the field is above 1.5 T. The energy cost by orientating magnetic moment from easy to hard axes, namely, the anisotropy constant, K = 1.84 × 10 6 erg/cm 3 at 10 K, is calculated from the area encircled by the two M-H curves. For the bare LSMO film, however, K = −0.93 × 10 6 erg/cm 3 at 10 K. Obviously, the actual anisotropy constant of the trilayers should be larger than that observed here since it has to overcome the intrinsic anisotropy energy of the bare LSMO. A simple calculation yields ΔK = 2.77 × 10 6 erg/cm 3 for trilayers. This value is larger by a factor of three in magnitude than that of the bare LSMO layer. Obviously, the effect of interface reconstruction is much stronger than the strain effect.
Based on Eq. (1), we can get an estimation of anisotropic energy. For LSMO, Mn possesses a spin-orbital coupling coefficient of ζ = 0.045 eV 43 and an orbital magnetic moment of ∼0.1 μ B . 44 Based on the density functional theory calculations, the orbital magnetic moment difference along two orthogonal directions can be up to Δμ L = 0.01 μ B . 45 Adopting this Δμ L , the calculated anisotropy constant is ∼10 6 erg/cm 3 , which is consistent with the experimental value obtained for our LCO/LSMO heterostructure.
Following a similar procedure, the anisotropy constant at high temperatures can be obtained. As shown in Fig. 6(d), K is maximal at low temperatures, nearly constant from 10 K to 100 K, and rapidly decreases upon further warming up above 100 K. The decrease in K with temperature can be ascribed to the decrease in the magnetization of the LSMO layer.
If the anomalous magnetic anisotropy stems from interface reconstruction, it should be dependent on the layer thickness of LSMO (t LSMO ). In Figs. 7(a)-7(c), we show the isothermal magnetization curves of the trilayers with different LSMO layers (see Fig. S5 of the supplementary material for M-T curves measured under different applied fields). From first glance, the discrepancy of the two M-H curves along two directions reduces as layer thickness grows. This feature is especially obvious as t LSMO increases from 10 nm to 19 nm. It implies a reduction in anisotropy energy. A direct calculation shows that K is ∼2.55 × 10 6 erg/cm 3 for t LSMO = 5 nm and ∼0.6 × 10 6 erg/cm 3 for t LSMO = 19 nm. The maximal anisotropy constant is 5.8 × 10 6 erg/cm 3 , gained in the [LCO (3 m)/LSMO (3 m)]5 superlattice. Figure 7(d) shows the anisotropy constant as a function of the reciprocal layer thickness of LSMO. Although the K-1/t LSMO relation is nonlinear, K displays a clear tendency toward growth as t LSMO decreases. This result confirms the dominative role of the interface effect. 46 Finally, we emphasize that the effect of interface reconstruction on spin orientation is strong only for tensile trilayers. For compressive trilayers, charge transfer takes place via both d 3z 2 -r 2 and d x 2 -y 2 orbitals. In this case, which orbital is more stable depends on the competition of these two orbitals. Consequently, the interface effect on the spin degree of freedom could be weakened. To investigate the effect of interface coupling in the

CONCLUSIONS
In conclusion, a systematic investigation on tensely strained (110)-LaCoO 3 /La 2/3 Sr 1/3 MnO 3 /LaCoO 3 trilayers is presented, focusing on the orbital reconstruction at interfaces and the effect stemming from spin-orbital coupling. It is found that the interlayer coupling makes Mn 3d x 2 -y 2 preferentially occupied, overcoming the effect of tensile strains which stabilizes d 3z 2 -r 2 . We present evidences for interlayer charge transfer via d x 2 -y 2 orbitals. This causes the formation of Mn-O-Co covalence, thus lowering the energy level of d x 2 -y 2 . In response to orbital reconstruction, the spin orientation of La 2/3 Sr 1/3 MnO 3 undergoes a 90 ○ switching in the film plane due to spin, charge, and orbital coupling. This work demonstrates how spin, charge, and orbital degrees of freedom couple with each other during the interface reconstruction, paving the way toward the exploration for novel materials. There are no conflicts of interest to declare.