Introduction

The asymmetry at the heterostructure interface of 3d transitional metal ABO3 oxides, including the mismatch of lattice constant, oxygen octahedral rotation and distortion, and chemical environment, has profound influences on spin and orbital coupling, yielding emerging phenomena such as enhanced ordering temperature, induced interfacial magnetism at the superconductor/manganite interface and orbital reconstruction1,2,3,4,5,6,7,8,9,10,11. Coupling between crystal structures12,13,14,15,16,17,18 across the heterostructure interface leads to new properties, a film-thickness-dependent interfacial ferromagnetic-polaronic insulator in Pr0.67Sr0.33O3/SrTiO319, spontaneous deep polarization15 in SrTiO3 near the La2/3Sr1/3MnO3/SrTiO3 interface, rotation of the magnetic easy axis of La2/3Sr1/3MnO320 and SrRuO321 thin film. Most work focuses mainly on the crystal structure without explicitly considering the effect of charge transfer. In 3d oxides, the properties of transition metal oxides are sensitive to electronic structure, especially the occupation of 3d orbitals4, which may be manipulated by doping with different covalent ions22, hydraulic pressure23, and external electric bias24,25. The chemical asymmetry, such as polar discontinuity, of the ABO3 perovskite heterostructure interface may be exploited to study the effect of charge transfer and crystal structure26,27,28.

Recent work13 claimed that oxygen octahedral coupling (OOC) is the leading factor affecting orbital hybridization in ultrathin manganite film; however, the effect of charge transfer on magnetic properties is also important6,7,8,9,10,11. Here, we consider a model system—La0.67Sr0.33MnO3/SrTiO3 (LSMO/STO) superlattice deposited on an (001) STO substrate. Bulk LSMO with high Curie temperature and high spin polarization28 has a rhombohedral unit cell29 with space group R-3c (No. 167) and lattice constants ar = 5.471 Å and αr = 60.43° at room temperature. The calculated pseudocubic lattice constant is 3.880 Å, and the MnO6 rotation pattern is a-a-a- with rotation angle 2.2°. STO is cubic30 at room temperature with lattice constant ~3.905 Å close to the lattice constant of LSMO, without TiO6 rotation. For the (001)-oriented heterostructure, the TiO2 and SrO atomic planes of STO are charge-neutral (no planar charge), while MnO2 and La/SrO atomic planes of LSMO have negative and positive planar charge, respectively, as illustrated as a blue bar at right bottom of Fig. 1a. The polar discontinuity exists around the LSMO/STO heterostructure, and the charge transfer across the interface occurs to avoid the polar catastrophe31,32. The final distribution of planar charge is illustrated as a pink bar at the right bottom of Fig. 1a, and the short-range charge transfer is constrained to several unit cells (UCs) near the interface32,33,34,35, illustrated as the probability of charge transfer at the left bottom of Fig. 1a. The superlattice with multiple heterostructure interfaces provides a viable way to manipulate the intensity of interfacial coupling, which may facilitate the short-range charge transfer.

Fig. 1: Magnetic properties of STO/LSMO superlattice films by SQUID.
figure 1

a Left: Illustration of probability of electronic charge transfer at STO/LSMO interfaces. The probability is normalized to the probability at the interface. The red curve is the probability of accepting an extra electron from LSMO across the interface by STO (e-acceptor), and the blue curve is the probability of losing electron to STO across the interface from LSMO (e-donor). Right: The illustration of planar charge distribution before (blue bar) and after (pink bar) the charge transfer. b The magnetization-temperature (MT) curves for SLn, measured with 100 Oe field. The inset is the derivative of the MT curve of SL2. The magnetic hysteresis loop at TRe = 0.87 for c SL10 and d SL2. The inset in each figure illustrates the effective anisotropy due to the sum of the in-plane and the out-of-plane magnetic anisotropy. e The summary of magnetic anisotropy for SLn at different temperatures, calculated from loops along the [001] and [100] directions. The summary of deconvoluted anisotropies for SLn at f TRe = 0.87 and g TRe = 0.05, respectively. The pink horizontal dashed line on the right indicates the Hb, −524 Oe in (f) and −5290 Oe in (g), of 60 unit cells LSMO layer on STO substrate

Full size image

In this work, the interfacial coupling of (LSMO)n/(STO)n superlattices (SLs) is manipulated by controlling the periodic thickness with n UCs of STO and n UCs of LSMO with a fixed thickness of ~120 UCs. The easy axis of magnetic anisotropy rotates ~45° towards the out-of-plane direction from n = 10 to n = 2 at reduced temperature TRe = T/TS = 0.87, where TS is the onset point of magnetization. Orbital hybridization accompanying charge transfer from Mn 3d and Ti 3d orbitals results in preferred occupancy of the \(3d_{3z^2-r^2}\) orbital at the interface and induces a stronger electronic hopping integral and interfacial magnetic anisotropy along the out-of-plane direction. As a result, the rotation to the out-of-plane direction of the effective magnetic easy axis for n = 2 is observed. This work reveals the predominant effect of charge transfer on the orbital hybridization and interfacial magnetic anisotropy.

Materials and methods

[(LSMO)n/(STO)n]m SLs with different periodic thickness (n UC LSMO and n UC STO, SLn), as illustrated in Fig. 1a, were grown on the (001) STO substrate by pulsed laser deposition at 100 mTorr oxygen pressure with the substrate temperature at 780 °C. Before film deposition, the SrTiO3 substrate was treated with a buffered NH4F-HF solution, then annealed at 975 °C for 2 h to obtain TiO2 plane termination. The reflection high-energy electron diffraction (RHEED) was used to monitor the deposition rate. The energy density of the 248 nm KrF excimer laser beam on target was 1.5 J/cm2 at a pulse frequency of 2 Hz. After deposition, the sample was cooled at 15 °C/min in a 200-mTorr oxygen atmosphere. The top surface is n UC STO, and the total thickness was kept at 2n × m = 120 UC (except for n = 8, for which the total thickness is 128 UCs). The final surface morphology with roughness ~0.3 nm was measured by atomic force microscopy (Fig. S1). The crystallographic properties of the films at room temperature were studied using a four-circle diffractometer (Huber 4-circle system 90000–0216/0) at the Singapore Synchrotron Light Source (SSLS), with an X-ray wavelength equivalent to Cu Kα1 radiation. The magnetic properties were measured by a superconducting quantum interference device (SQUID), and the diamagnetic signal from the STO substrate was subtracted to obtain the magnetic signal from LSMO/STO superlattice only. The magnetization value is calculated by \(M/\left( {\frac{A}{{a^2}}t} \right)\), where M is the measured magnetic moment, A is the in-plane area of the sample, a is the in-plane lattice constant (which is the same as the in-plane lattice constant of the STO substrate), and t is the thickness of the LSMO of the whole sample in number of UCs (64 for SL8, and 60 for all other samples). The transport properties were measured by the Physical Property Measurement System (PPMS) using a linear four-point probe. The X-ray absorption near edge structure (XANES) measurements were carried out using linear polarized X-rays at the undulator beamline 20-ID-B of the Advanced Photon Source (APS), Argonne National Laboratory (ANL), USA. The Si (111) monochromator with resolution δE/E = 1.3 × 10−4 was used. The XANES was collected in the fluorescence mode using a 12-element Ge solid-state detector. A Mn metal foil, placed to intercept a scattered beam, was used as an online check of the monochromator energy calibration. To get reliable information, each XANES curve was obtained by averaging the results of five measurements with total counts of more than one million. The X-ray magnetic circular dichroism (XMCD) with different incident angles at L3,2 edges of Mn was measured with a magnetic field of 5 kOe parallel to the wave vector of the X-ray, at 4-ID-C, APS in ANL and SINS beamline in SSLS, using the total electron mode for data collection at different temperatures. After the raw absorption spectra were normalized to the incoming photon intensity, a linear background obtained by fitting the absorption at the pre-edge range was subtracted, and the absorption at the pre-edge was at zero intensity. The absorption spectra were normalized to the post-edge of absorption. The normalized XMCD36 was obtained by (ρ+−ρ−)/(ρ+ + ρ−), where ρ+ and ρ− are absorption with the helicity parallel and antiparallel to the applied magnetic field, respectively. The first-principle simulation of X-ray absorption spectroscopy (XAS) and XMCD was conducted using the FDMNES code37,38, developed by Yves Joly, Yvonne Soldo-Olivier et al. at Institut Néel, France. The superlattice structure was built based on experimental results. The final states of the electrons in the cluster with radius of 6 Å were calculated using the multiple-scattering approach with spin-orbit coupling included, and the chemical valence of Mn was controlled by setting the number of 3d electrons accordingly. To account for the core-hole lifetime, the calculated XAS and XMCD spectra at Mn L edges were convoluted with a Lorentzian function. In addition, to account for the experimental resolution and scaling of the spectra, a complementary convolution with a Gaussian function with typical broadening parameters of σ = 1.0 eV was performed. The spectra of Mn with various chemical valences are normalized for easy comparison. Select area electron diffraction (SAED), high resolution TEM (HRTEM), scanning transmission electron microscopy-high angle annular dark field (STEM-HAADF), and electron energy loss spectroscopy (EELS) spectrum imaging were performed using the double aberration-corrected JEOL-ARM200CF microscope with a cold-field emission gun and operated at 200 kV, at Brookhaven National Laboratory, USA. The microscope is equipped with JEOL and Gatan HAADF detectors for incoherent HAADF (Z-contrast) imaging, Gatan GIF Quantum ER Energy Filter with dual EELS for EELS. The cross-section TEM samples were prepared by focused ion beam (FIB) followed by gentle ion-milling at low operating voltages. The STEM-HAADF images were acquired with a condensed aperture of 21.2 mrad and collection angle of 67–275 mrad.

Results and discussion

The Magnetization-Temperature (MT) curves in Fig. 1b show that the paramagnetic-to-ferromagnetic phase transition temperature decreases with decreasing n. In the double-exchange model22, ferromagnetic exchange between Mn3+ and Mn4+ ions occurs via the oxygen 2p orbital. An electron from Mn3+ transfers to the oxygen simultaneously with oxygen transferring an electron to Mn4+. This transfer effectively facilitates the movement of spin from a Mn3+ ion to a Mn4+ ion without spin flip. The delocalization of the electron reduces the kinetic energy, thus enabling the ferromagnetic exchange. The intensity of this double-exchange interaction depends on the local atomic and chemical environment. At the LSMO/STO interfaces, the magnetic exchange interaction is weakened, which is observed when the n decreases (the number of LSMO/STO interfaces increases) the temperature of phase transition of the LSMO decreases. The inset of Fig. 1b shows the broad magnetization transition for SL2, which does not allow for an unambiguous definition of the Curie temperature. Then, a different reference temperature TS is introduced for analysis, defined as the temperature at which magnetization of the material starts to appear. The derivative of MT curve shows that the magnetic moment starts to increase at TS = 172 K in SL2. At the beginning of the phase transition, a cluster or spin glass state may exist due to the spin frustration at LSMO/STO interface, to be`discussed later. To compare magnetic anisotropy between different samples, the reduced temperature TRe = T/TS is employed. The summary of TS and TRe is shown in Table S1 for different samples. At TRe = 0.87, Fig. 1c, d shows that the out-of-plane [001] axis of SL2 is easier to saturate than the in-plane [100] axis, which is opposite for SL10. These hysteresis loops30 clearly show a change in magnetic easy axis direction between SL2 and SL10. See Fig. S2 for more hysteresis loops. The saturated magnetization of bulk LSMO22 is 3.7 μB/uc, higher than the saturated magnetization of the film in the current work. The weakening of magnetic properties of manganite thin films has previously been observed39. The decrease in the saturated magnetization with n suggest that the effect of the LSMO/STO interfaces on the magnetic properties is significant. The hysteresis loop along the [110] is quite close to that along the [100] direction in Fig. 1c, d, and there is a weak in-plane anisotropy with the easy axis along the [110] direction. Similar in-plane weak anisotropy has been reported in a single LSMO layer on the [001] STO substrate39. From the area enclosed by the in-plane and the out-of-plane hysteresis loops (effective magnetic anisotropy energy, Keff) in the first quadrant, the effective magnetic anisotropy could be calculated via Heff = −2Keff/μ0MS, where Heff is the effective magnetic anisotropy field, and MS is the saturated magnetization. For SL2, the difficulty of saturating the material along different directions follows the sequence: [001] << [110] < [100]. Although for SL10 the sequence of difficulty to saturate the material is [110] < [100] << [001], the energy difference required to saturate the sample between [110] and [100] directions is quite small, as shown in Fig. 1c. Then, magnetic anisotropy between [001] and [100] directions is discussed, unless specified. The small remnant of magnetization in SL2 depends on the magnetic domain structure and the magnetization process40, which is not in the scope of this work.

Figure 1e summarizes Heff for all SLs at different temperatures. The positive value of H denotes an easy axis normal to the film plane, and the negative value denotes an easy axis in the film plane. With increasing temperature and decreasing n, the magnetic anisotropy favoring in-plane decreases gradually (negative Heff approaching zero) for all samples, and SL2 shows a magnetic easy axis favoring the out-of-plane direction at high temperature TRe = 0.87 with positive Heff. Heff could be viewed as the sum of terms favoring the in-plane orientation (H//) and that favoring the out-of-plane orientation (H⊥), as illustrated in the inset of Fig. 1c, d. The effective anisotropy field41 is deconvoluted as Heff = Hi + Hb − Hd, including bulk term Hb and demagnetization term Hd favoring the in-plane direction, and interfacial term Hi favoring the out-of-plane direction. The demagnetization term could be calculated by Hd = −N·MS, while the demagnetization factor is N = 1 along the out-of-plane direction for thin films. Hb + Hi is the difference between Heff and Hd. For different samples, all the growth parameters such as temperature, oxygen pressure, and laser energy are the same, and the only difference is the number of LSMO/STO interfaces. Any change in Hb + Hi with n should therefore come from the LSMO/STO interfacial coupling, and the bulk term Hb should be constant for different samples. In this work, the Hb is obtained from a reference sample of 60 UC single layer LSMO on the (001) STO substrate (without depositing any STO layers), the interfacial term Hi of which is ignored. Due to the multiple contributions to Heff, this parameter experiences a change in direction of magnetic easy axis with n and temperature. Figure 1f shows the trend of deconvoluted anisotropies with n at TRe = 0.87 and the constant bulk Hb term (dashed horizontal pink line). The demagnetization effect Hd decreases with decreasing n due to decreasing MS. For n = 2, Hd is close to zero due to the low magnetization MS = 0.11 μB/uc. Hi increases with decreasing n, which, complementary to the decreased Hd, induces rotation of the easy axis towards the out-of-plane direction in SL2. Current experimental results reveal that multiple interfaces in SLs render Hi non-trivial.

With decreasing temperature, the magnetic anisotropy favors the in-plane direction for all samples. Taking TRe = 0.05 as an example, Fig. 1g shows the trend of deconvoluted anisotropies with n. Hi increases with decreasing n, following the similar trend at high temperatures (Fig. 1f). Compared to high temperature TRe = 0.87, at low temperatures TRe = 0.05 Hi increases slightly. However, the bulk term Hb and the demagnetization term Hd are significantly enhanced. As a result, the effective anisotropy Heff favors the in-plane direction at TRe = 0.05. In addition to rotating the direction of the easy axis, the interfacial coupling could induce a certain degree of spin frustration around the interface42, as revealed by the magnetic memory effect as shown in Fig. S3. The memory effect is obvious in SL2 at 15 K but is relatively weak in SL6 and SL10. These results suggest that the spin frustration may be related to the LSMO/STO interfacial coupling, i.e., a higher number of LSMO/STO interfaces promotes stronger spin frustration. The enhanced spin frustration in SL2 may contribute to the broad magnetization transition as shown in Fig. 1b. The above discussion reveals that the interracial coupling exists in a wide temperature range.

Angularly dependent Mn L3,2 edge XMCD yields element-specific information of magnetic anisotropy at different temperatures, as shown in Fig. 2 and Fig. S4. The incidence angle θ = 0° corresponds to the wave vector of the X-ray along the normal direction of the film plane, and the magnetic field is always parallel to the wave vector. The probing depth of XMCD is governed by the electron escape depth (λe, several nm), independent of the incident angles for the same sample. A simplified estimation of the saturation effect43 of angularly dependent XMCD has been corrected by a factor \(f = \frac{1}{{1 + \lambda _{^e}/(\lambda _{^x}{\mathrm{\cos\theta }})}}\). λe is in the range of several nm44, and λe = 5 nm45 is an appropriate estimation for calculation in manganite; λxcosθ is the X-ray penetration depth. λx, X-ray penetration length, around the Mn L absorption edge is estimated to be ~198 nm according to the previous report46, ignoring the energy dependence of λx as discussed in ref. 47. Then, the magnetic easy axis of a particular sample is identified by the maximum XMCD intensity at different incident angles. Figure 2c shows the summary of XMCD intensity at the L3 edge, ~639 eV for SL2. With decreasing temperature Tre from 0.87 to ~0.05, the easy axis rotates from 15° to 60° for SL2. However, the easy axis does not change for SL10 (60°) in the temperature range shown in Fig. 2f. At the same reduced temperature Tre = 0.87, the direction with maximum XMCD intensity is θ = 60°, 45°, and 15° for n = 10, 6, 2, respectively. In SL10, the LSMO/STO interfacial coupling still exists, including the coupling of lattice constant and charge transfer as discussed later. The interfacial contribution Hi may cause the easy axis at 60° with respect to the surface normal rather than being completely in the in-plane direction. Compared to SL10, the easy axis of magnetic Mn ions in SL2 rotates 45° towards the out-of-plane direction, which is consistent with the SQUID results. The maximum intensity at 60° for SL10 at low temperature Tre = 0.03 in Fig. 2b shows the same trend of establishment of the magnetic easy axis towards the in-plane direction. Angular dependent XMCD therefore indicates that Mn ions have an important contribution to the rotation of magnetic anisotropy in SLn. The contribution from the Ti sites to the total magnetic moment is insignificant as calculated by the sum rules48 from Ti L3, 2 XMCD (Fig. S5). Hence, the discussion of magnetic properties focuses on Mn sites hereafter.

Fig. 2: Angular Mn L edge XMCD for SL2, SL6 and SL10 STO/LSMO superlattice films.
figure 2

Angular dependent Mn L3,2 edge XMCD for SL2 (a, b), and SL10 (d, e) at different temperatures. The XMCD intensity at 639 eV is summarized in (c) to determine the easy axis of the SL2 at different temperatures. The curves for different temperatures are vertically shifted for comparison. f The summary of the incident angle θ with maximum XMCD intensity for SL2, SL6, and SL10 at different temperatures. Inset of (f) is the setup for the measurement

Full size image

Both the SQUID and angularly dependent XMCD reveal an out-of-plane magnetic easy axis for SL2 at Tre = 0.87 due to the strong interfacial contribution. The effect of interfacial OOC on magnetic anisotropy has been investigated recently13,20,21,49. In this work, quantitative analysis of the crystal structure using X-ray diffraction, half-integer diffraction50,51,52 (Fig. S6, and Table S2), detecting three-dimensional MnO6 rotation, shows a large difference of the Mn–O bond angle and bond length along the in-plane and the out-of-plane directions in SL10. In SL2, although the measured bond lengths of both directions are similar, the larger in-plane bond angle (Fig. S6B) suggests a higher in-plane hopping integral, and the magnetic moment would favor the [100] axis instead of the [001] axis, as observed. ref. 13 claimed that OOC is the leading factor affecting orbital hybridization to determine material properties. However, octahedral rotation cannot explain the results of the current LSMO/STO superlattice. Other possible factors include the strong 3d spin–5d orbital coupling53, as proposed in the La2/3Sr1/3MnO3/SrIrO3 interface, which is relatively weak between the 3d transition metals (Mn and Ti) in the LSMO/STO superlattice. Therefore, it is necessary to consider other factors responsible for the observed phenomena. Then, high resolution scanning transmission electron microscopy (STEM) was used to probe the local structural information at the LSMO/STO interfaces for n = 2, 6, 10 in Fig. 3. The STO substrate, at the left side of the image, was treated with a TiO2 termination before deposition, and atomically flat LSMO/STO interfaces existed in these samples.

Fig. 3: HAADF-STEM images of (a–c) SL10, SL6, and SL2 STO/LSMO superlattice films.
figure 3

The out-of-plane plane spacing cint (blue squares) and Ti chemical valence (red dots) as a function of position are shown in the middle and bottom part of a–c. The horizontal orange dashed lines mark the cint of STO in the bulk substrate. d Ti EELS at L3,2 edges from SrTiO3 substrate far from interface (I: red line), sample surface (II: black line), interface between LSMO/STO superlattice (III: orange circles), center of deposited STO layer in LSMO/STO superlattice (IV: olive circles) in SL6, and deposited STO layer in SL2 film (V: blue circles). The spectra of I and II are close to those of bulk SrTiO3 and LaTiO3, respectively and are thus used as the reference spectra of Ti4+ and Ti3+. The appearance of Ti3+ is due to the lack of oxygen coordination at the sample surface53. Then, the fitted Ti valence is Ti3.97+, Ti3.78+ and Ti3.75+ for III, VI and V, respectively. e, f High resolution TEM (HRTEM) image for SL6 and SL2. The second and third rows correspond to the relative change in the out-of-plane (second row) and the in-plane (third row) spacing of SrO (La/SrO) planes. The vertical white dashed lines at two STO/LSMO interfaces are for eye guidance

Full size image

By refining the peak positions, the (001) plane spacing of two successive La/SrO (or SrO) planes along the [001] direction (cint) was obtained. For SL10 and SL6, the cint shows an oscillating pattern with a maximum value at the left STO/LSMO interface of each LSMO layer, larger than 3.905 Å, and with a minimum at the right LSMO/STO interface of each LSMO layer, which indicates an asymmetric coupling at two interfaces of each LSMO layer and is related to the growth process (to be discussed later). The oscillation pattern of cint in film is further confirmed by geometric phase analysis (GPA)54,55 from the HRTEM image (the second row of Fig. 3e). There is basically no change in the in-plane lattice (the third row in Fig. 3e) over the whole image except some random changes caused by the image noise. For SL2, there is no obvious pattern of cint, but cint of the whole film is slightly larger (cint/a > 1) than cint of the substrate STO, revealed as the dominant red part of GPA (second raw of Fig. 3f). The SAED is used to separate the contribution of the diffraction peak between the film and the substrate in Fig. S7. The 002 diffraction peak of SL2 film (blue line) clearly shifts to the left with respect to the substrate (red line), confirming that the cint of the SL2 film is slightly larger than the cint of the STO substrate. The SAED reveals that a = 3.905 Å, and cint = 3.911 Å for SL2, which is consistent with the XRD results in Fig. S8. The shoulder at the right side of the 002 X-ray diffraction reflection, corresponding to the film, is quite clear for n ≥ 6, but disappears for n < 4, revealing that the out-of-plane lattice constant increases with the decrease in n for SLn. A similar enlarged tetragonal ratio around the interface has been reported in other systems56, which may affect the electronic structure. In addition, the Mn displacement from the body center of the pseudocubic UC, recently reported by Guo et al.57, is observed and discussed in Fig. S9.

Volume conservation is generally not observed in epitaxial thin film58, due to the strain effect of the substrate or interfacial effects (such as chemical interdiffusion and charge transfer). In the case when the lattice constant is dominated by the substrate-induced strain effect58, the in-plane tensile strain normally induces out-of-plane compressive strain with c/a < 1. However, in our current work, the LSMO is under in-plane tensile strain with a = 3.905 Å, and the out-of-plane lattice around the LSMO/STO interface is larger than the in-plane lattice constant (cint = 3.911 Å for SL2) with c/a > 1. Hence, other factors should be present. Inspired by a high correlation between crystal and electronic structures in the 3d oxide, the electronic charge distribution was examined by Ti L edge EELS34,59,60, summarized in Fig. 3a–d. SL10 and SL6 show the same trend of Ti chemical valence. Taking SL6 as an example, the chemical valence of Ti at the LSMO/STO interface (~3.78) decreases compared to the chemical valence of Ti at the center of the STO layer (~3.97), indicating that extra electrons are accepted by the Ti 3d orbitals at the LSMO/STO interface. The critical thickness of the charge transfer from the STEM results is ~3 UCs, consistent with the observations of others34. The averaged chemical valence of Ti is ~3.75 in SL2 with each STO UC situated at the LSMO/STO interface. There is no clear trend of peak position of Mn61,62 in Fig. S10, similar to ref. 13. However, the shape of the absorption different for Mn with a different chemical valence, even though change in peak position may be too subtle to detect. The first principle XAS simulation was then carried out using FDMNES37,38, which is a method different from the configuration interaction model63,64 and could be used to investigate the effect of the heterostructure interface. The comparison of the shape of the X-ray absorption curve between the simulated and measured Mn L edge XAS in S10 suggests Mn chemical valence higher than the bulk value of 3.3. The EELS shows evidence of charge transfer from Mn to Ti 3d orbitals across the LSMO/STO interface, and the probability is the highest for SL2 (lowest Ti valence) with the maximum number of interfaces. One other possible reason for the change in the Ti chemical valence is the oxygen stoichiometry around the LSMO/STO interface. In this work, the LSMO and STO layers for all samples were fabricated using the same growth parameters, including the oxygen pressure. If the change in Ti valence is from oxygen stoichiometry, the averaged Ti chemical valence should be the same for all samples, while SL2 has the lowest averaged Ti chemical valence compared to others. Hence, the dominant effect on the change in the Ti chemical valence should arise from the electronic charge transfer across the LSMO/STO interface.

In addition, Mn L edge X-ray linear dichroism (XLD) was performed to study the electronic configuration, as shown in Fig. 4a, b. The parallel XLD (I||) was measured with the wave vector of incident X-ray parallel to the normal direction of the film plane, and the electric field of the X-ray is parallel to the film plane. The perpendicular XLD (I) was measured with the wave vector of incident X-ray 70° to the normal direction, and the electric field of the X-ray is 70° to the film plane. The XLD was calculated as the intensity difference (I||-I) between the spectra measured with in-plane (E||) and out-of-plane (E) polarizations. The XLD was measured at 300 K for n = 2, 4, 6, and 350 K for n = 10. The difference of two X-ray absorption spectra shows the difference of occupancy in the two eg orbitals. The gradual decrease in intensity of the feature at ~ 640 eV is due to the gradual increase in relative occupancy of \({3d}_{3z^2-r^2}\) orbitals with decreasing n, as shown in Fig. 4b; and SL2 shows the preferred occupancy of \({3d}_{3z^2-r^2}\) orbitals65,66. The resultant depletion of the \({3d}_{3z^2-r^2}\) orbitals weakens the ferromagnetic double-exchange interaction65 and lowers the paramagnetic-to-ferromagnetic phase transition temperature in SL2.

Fig. 4: Orbital occupancy and charge transfer in STO/LSMO superlattice films.
figure 4

a The detailed XAS curves for SL2, and the summarized Mn L edge XLD for four samples is shown in (b). The curves for different samples are vertically shifted for comparison. c The effect of the magnetic field on the angular dependence of magnetoresistance (AMR) of SL3 at TRe = 0.04 (10 K). All curves are normalized to the resistance at θ = 0°. The measurement configuration is illustrated as the inset of (c), and the current is along the (100) direction; d Summary of AMR at 10 K with magnetic field of 80 kOe for different samples: TRe = 0.04 for SL3, SL4 and SL6, and TRe = 0.03 for SL10

Full size image

The charge transfer with corresponding orbital occupancy has a strong effect on angular-dependent magnetoresistance (AMR) at the low temperature range. During AMR measurement, the magnetic field rotates from the out-of-plane to the in-plane direction and remains perpendicular to the current (along [100] direction) as shown in the inset of Fig. 4c. Around these temperature ranges, all films show in-plane magnetic anisotropy (Fig. S2). Normally the resistance minimum should occur with the magnetic field in the film plane (θ = 90°). However, with increasing magnetic field from 1 to 80 kOe, the resistance minimum point SL3 changes from the in-plane to the out-of-plane (θ = 0°) directions. This phenomenon is related to orbital reconstruction67 with charge transfer across the LSMO/STO interface, especially the \({3d}_{3z^2-r^2}\) orbital occupancy, and the AMR increases with decreasing temperature67. When the magnetic field is in the film plane, the \({3d}_{3z^2-r^2}\) band flattens at the Fermi level due to the Mn spin-orbit coupling and results in a larger resistivity compared to the resistivity with the magnetic field along the out-of-plane direction due to the decreasing electronic velocity. Figure 4d shows that the AMR effect is largest for SL3, and the AMR effect decreases with increasing n. SL10 has a smaller AMR at the reduced temperature TRe = 0.03 compared to others at TRe = 0.04, suggesting that SL10 has the lowest AMR effect at the same reduced temperature among these SLs. Although the resistance of SL2 exceeds the low temperature measurement capability as shown in Fig. S11, the trend of AMR indicates that the effect of orbital reconstruction at the LSMO/STO interface increases with decreasing n (enhanced interfacial coupling). Additional experimental results show that AMR with the currents along both [100] and [110] directions shows the same trend as discussed above. The large difference between AMR in SL4 with currents along the [100] and [110] directions may be related to the anisotropy in the space of electron density68, or the local anisotropy of spin-orbital coupling69 with orbital reconstruction at the LSMO/STO interface. The exact origin of the dependence of AMR on the current direction warrants further future work. See S11 for more details.

The origin of the preferred occupancy of \({3d}_{3z^2-r^2}\) at the heterostructure interface was attributed to arguments such as the interfacial symmetry breaking and new crystal structure phases53,62. Based on the existence of the charge transfer and AMR, a mechanism due to orbital hybridization with charge transfer across the interface70 is proposed (Fig. 5a). From the perspective of molecular orbitals, the interfacial bonding lowers the energy of 3d orbitals, favoring the occupancy of the \({3d}_{3z^2-r^2}\) orbital due to a large spatial overlap. When this interfacial orbital forms, the electronegativity of interfacial bonds may facilitate the charge transfer between Mn and Ti 3d orbitals. With the correlation between crystal structure and electronic structure in 3d oxides, the preferred \({3d}_{3z^2-r^2}\) orbital occupancy may enlarge the out-of-plane cint to lower the total energy, as supported by the enlarged cint/a > 1 around the left LSMO/STO interface of each LSMO layer (Fig. 3). The oscillating pattern of cint for SL10 and SL6 (Fig. 3a, b) could be understood as follows. For one LSMO layer, the 1st UC LSMO at the left has a strong coupling with its left STO layer. Thus, the 1st UC LSMO has a large cint with charge transfer. The coupling decreases with increasing thickness of LSMO and cint/a decreases to <1. Along with the further increase in thickness with negligible charge transfer, cint continues to decrease due to in-plane tensile strain and reaches a minimum value at the right interface. At the right interface, the cint/a < 1 even with the existence of charge transfer. The subsequent 1st UC STO couples to the preceding LSMO layer with minimum cint; with increasing thickness of STO, the cint increases until reaching the next STO/LSMO interface.

Fig. 5: Anisotropic hopping integral in STO/LSMO superlattice films.
figure 5

a Illustration of the 3d orbital energy levels after hybridization between Mn and Ti across the interface. More discussion in the text. b The illustration of electronic hopping along the in-plane integral (tin) through Mn 3d\(x^2-y^2\), and the out-of-plane (tout) direction through Mn \({3d}_{3z^2-r^2}\). c The illustration of electronic configuration before and after X-ray absorption. The L denotes the six oxygen ligand around Mn, and L denotes one electron missing from the ligand, which is located at the Mn site orbital orbitals through hopping between O 2px or 2py and Mn 3d\(x^2-y^2\) orbital. Similar hopping would happen between O 2pz and Mn \({3d}_{3z^2-r^2}\) orbitals. See refs. 45, 64, 65 for more details. d The derivative of the polarization-dependent Mn K edge XANES for SL2 with the electric field of the X-ray parallel (||) and perpendicular (⊥) to the film plane. There are two features (B1 and B2) at the K edge. The enhanced B1 peak in the parallel configuration indicates the stronger electronic hopping integral along the out-of-plane direction (tout > tin). e The comparison of XAS for different samples in parallel configuration, which relates to the out-of-plane electronic hopping integral tout. The lower inset is the enlarged part around B1 and B2 peaks; the upper inset indicates the trend of Λ||/Λ⊥. Λ|| and Λ⊥ indicate the IB1/IB2 ratio in parallel and perpendicular configurations

Full size image

The above results suggest strong coupling around the LSMO/STO interface, and the existence of multiple LSMO/STO interfaces facilitates the charge transfer. With the increase in the interface number (decreasing n), the interfacial term Hi becomes more dominant. The trend of Hi could be addressed from the electronic hopping integral, as below. For a tight-binding Hamiltonian20 of the LSMO ultrathin film, \(\sum _{\it{R}}t_{\alpha \beta }\left( R \right)e^{iK \cdot R} + \left( {\frac{\lambda }{2}} \right)\sigma \left( {\theta ,\varphi } \right) + \xi L \cdot S\) was proposed to explain the magnetic anisotropy energy (MAE) according to density functional theory. The tαβ (R) represents the hopping integral from orbital α at the original site to orbital β at site R. K is the waver vector, λ is the exchange splitting, the magnetization is along the (θ, φ) direction, and σ(θ, φ) is the vector of Pauli matrices times a unit vector in the direction (θ, φ). ξ represents the spin (S)-orbital (L) coupling intensity. The structural change mainly affects tαβ (R), which in turn leads to the change in magnetic anisotropy. A high hopping integral is favored along the magnetic easy axis in a ferromagnetically ordered state. In the current work, the hopping integral along the in-plane tin and the out-of-plane tout direction is illustrated in Fig. 5b. The relative intensity of the hopping integral correlates with the detailed shape of the polarization-dependent Mn K edge XANES51,71,72 as shown in Fig. 5c–e and Fig. S12. The electronic configuration of the MnO6 octahedron in the B1 state is proportional to the hopping integral, which could be used to characterize the in-plane and the out-of-plane hopping integral: the higher intensity ratio (Λ|| = IB1/IB2) in the parallel measurement with the polarization vector (E vector of X-ray) parallel to the film plane suggests a higher out-of-plane hopping integral tout. For SL2, the out-of-plane hopping integral tout is higher than the in-plane hopping integral tin in Fig. 5d, induced by the enhanced \({3d}_{3z^2-r^2}\) orbital occupancy. With the increase in interface numbers (decreasing n), Λ|| increases and the Λ||/Λ⊥ ratio increases in Fig. 5e, indicating that the perpendicular contribution to the effective magnetic anisotropy increases with the increase in interface number (decreasing n), consistent with the trend of Hi with decreasing n, which causes ~45° rotation of effective magnetic anisotropy towards the out-of-plane direction from SL10 to SL2 at TRe = 0.87. Please note that current work does not contradict ref. 13, which proposed that the change in crystal structure-MnO6 rotation is the leading factor to affect orbital hybridization in a bilayer or trilayer system. Actually, current superlattice work also shows the change in crystal structure in terms of the lattice constant around the interface, and further study on electronic structure with charge transfer echoes with the change in lattice constant. The periodic thickness, number of interfaces and total thickness may play a role in the extent of the influence of the interfacial orbital hybridization on the out-of-plane magnetic anisotropy. These possible approaches to achieve a larger angular rotation in magnetic anisotropy warrant further investigation.

Conclusions

In summary, the variation in periodic thickness of the (La0.67Sr0.33MnO3)n/(SrTiO3)n SLs affects the interfacial atomic arrangement and induced charge transfer and orbital hybridization. Increasing the number of interfaces enhances orbital hybridization, resulting in interfacial electronic charge transfer from Mn 3d and Ti 3d orbitals and corresponding rotation of the magnetic easy axis towards the out-of-plane direction for n = 2 at TRe = 0.87. This work demonstrates a promising approach of the use of SLs to control interface-induced properties of strongly correlated oxides in the development of novel magnetoelectronic devices.