Fully Magnetically Polarized Ultrathin La0.8Sr0.2MnO3 Films

We report the observation of fully magnetically polarized ultrathin La0.8Sr0.2MnO3 films by using LaMnO3 and La0.45Sr0.55MnO3 buffer layers grown epitaxially on SrTiO3(001) substrates by molecular beam epitaxy. Specifically, we show that La0.8Sr0.2MnO3 films grown on 12-unit-cell LaMnO3 have bulk-like magnetic moments starting from a single unit cell thickness, while for the 15-unit-cell La0.45Sr0.55MnO3 buffer layer, the La0.8Sr0.2MnO3 transitions from an antiferromagnetic state to a fully spin-polarized ferromagnetic state at 4 unit cells. The magnetic results are confirmed by X-ray magnetic circular dichroism, while linear dichroic measurements carried out for the La0.8Sr0.2MnO3/La0.45Sr0.55MnO3 series show the presence of an orbital reorganization at the transition from the antiferromagnetic to ferromagnetic state corresponding to a change from a preferred in-plane orbital hole occupancy, characteristic of the A-type antiferromagnetic state of La0.45Sr0.55MnO3, to preferentially out of plane. We interpret our findings in terms of the different electronic charge transfers between the adjacent layers, confined to the unit cell in the case of insulating LaMnO3 and extended to a few unit cells in the case of conducting La0.45Sr0.55MnO3. Our work demonstrates an approach to growing ultrathin mixed-valence manganite films that are fully magnetically polarized from the single unit cell, paving the way to fully exploring the unique electronic properties of this class of strongly correlated oxide materials.


INTRODUCTION
The disruption of the atomic structure at the boundary between different materials often gives rise to the emergence of new phenomena that are characteristic of the interface region itself. 1,2he study of such interface phenomena is important to understanding the role of broken symmetries, electron exchange, and correlation effects to the electronic properties, but may also hold promise for new electronic devices based on purely interfacial effects.Critical for strong interfacial effects is the growth of sharp and well-defined interfaces separating materials with pristine electronic structures; this is because the presence of defects at the interface may strongly disturb the electronic properties of the component phases, for instance, making the interdiffused interface of a magnetic material nonmagnetic 3 or where the presence of charge traps screens the polarization of a ferroelectric interface. 4−10 For the particular case of the Pb(Zr 0.2 Ti 0.8 )O 3 /La 0.8 Sr 0.2 MnO 3 interface, the magnetoelectric coupling was demonstrated to originate from charge-driven modulation of the valency and spin configuration of the interfacial La 0.8 Sr 0.2 MnO 3 layer induced by charge screening of the ferroelectric polarization. 11−16 The observed changes in the magnetic moment of the La 0.8 Sr 0.2 MnO 3 layer are on the order of 20%, which is a large effect; however, given the interfacial character of the magnetoelectric coupling, the change in the magnetic moment is associated with only a few unit cells of the whole sample, and therefore, the total change is averaged out by the remainder of the magnetic film.−29 It is therefore important to devise solutions for obtaining ultrathin films with bulk-like or enhanced interfacial characteristics, including full magnetic and ferroelectric polarizations in ferromagnetic and ferroelectric systems, respectively.In addition, one also expects that the relative importance of the interfacial effect will increase as the thickness is reduced since the interface will then be a larger portion of the overall system.Hence, it is important that not only the interface is well-defined but also that the electronic properties of the materials remain robust with decreasing thickness, ideally down to the monolayer range. 3,4o overcome the degradation in the magnetic properties of ultrathin manganites, one strategy consists of decoupling the films from the nonmagnetic substrate by inserting a magnetically polarized buffer layer. 20,30For example, the introduction of 2 uc LaMnO 3 between a 120 nm thick La 1−x Sr x MnO 3 (x = 0.4) and the SrTiO 3 substrate leads to the disappearance of the dead layer as measured by second harmonic signal generation and ascribed to the interruption of charge transfer across the SrTiO 3 interface, which is otherwise responsible for hole doping of the La 0 .6 Sr 0 .4 MnO 3 film. 2 0In another instance, 1 uc La 1−x Ba x MnO 3 (x = 0.3) films sandwiched between two 3 uc SrRuO 3 layers are found to be fully magnetically polarized, an effect attributed to the presence of oxygen octahedral rotations induced by SrRuO 3 , leading to an orbital reconstruction and an enhancement of the interfacial magnetic properties. 31In another example, CaRu 1/2 Ti 1/2 O 3 used as an interlayer in La 2/3 Ca 1/3 MnO 3 superlattices has been shown to preserve the magnetic properties of the individual manganite films and to result in antiferromagnetic interlayer coupling. 32These studies highlight the importance of separating the manganite film from the supporting SrTiO 3 substrate and highlight the role of charge transfer, spin exchange, and octahedral tilting, together with orbital rearrangement, in the improvement of the magnetic properties toward bulk characteristics.
In this study, we explore the magnetic and transport properties of ultrathin La 0.8 Sr 0.2 MnO 3 films when a manganite buffer layer is introduced at the interface with the SrTiO 3 substrate.We employ two different types of buffer layers, a nominally antiferromagnetic insulating LaMnO 3 and an antiferromagnetic conducting La 0.45 Sr 0.55 MnO 3 , in order to disentangle the roles of spin and charge exchange at the interface.In the first case, our results show that the La 0.8 Sr 0.2 MnO 3 film develops robust magnetic properties, with a near-bulk-like critical temperature and magnetic moments that are independent of thickness starting at 1 uc.In the second case, an antiferromagnetic coupling to the buffer layer is observed for thicknesses up to 3 uc, while for larger thicknesses, the top La 0.8 Sr 0.2 MnO 3 layer transitions to a fully polarized ferromagnetic state.By probing the electronic band structure using X-ray absorption spectroscopy, we link the magnetic properties of La 0.8 Sr 0.2 MnO 3 to the evolution of orbital occupancy with film thickness.

EXPERIMENTAL SECTION
The samples in this study consist of La 0.8 Sr 0.2 MnO 3 films with thicknesses ranging from 0 to 10 uc, grown on a 12 uc LaMnO 3 buffer layer (except for the 3 uc La 0.8 Sr 0.2 MnO 3 sample, grown on 14 uc LaMnO 3 ) or on a 15 uc La 0.45 Sr 0.55 MnO 3 buffer layer.Figure 1e shows a schematic diagram of the sample structure.The buffer layer thickness is chosen based on previous works showing that 12 uc LaMnO 3 is insulating and has a reduced magnetic moment 33 and that 15 uc La 0.45 Sr 0.55 MnO 3 is conducting antiferromagnetic. 34The samples were grown in an oxide molecular beam epitaxy (MBE) system with a base pressure of 2 × 10 −10 mbar, equipped with five effusion cells, a reflection high-energy electron diffraction (RHEED) system, and a load-lock chamber for sample transfer.For this study, we used commercially available SrTiO 3 (001) substrates (SurfaceNet), with ±0.1°miscut, which were chemically treated with a HCl:HNO 3 solution and annealed to 1000 °C in air to provide single TiO 2 -terminated surfaces. 35,36For metal oxide deposition, we set the molecular oxygen partial pressure to (5 ± 1) × 10 −7 mbar, the substrate temperature to 720 °C, and the evaporation rate to about 0.5 uc/min, values that we found to provide the optimal growth conditions.In all cases, we adopted the same procedure for preparing the SrTiO 3 substrate in situ, which consisted of heating slowly the substrate up to the growth temperature so as to keep the pressure below 2 × 10 −7 mbar and below 2 × 10 −9 mbar before growth.The surface of the SrTiO 3 (001) substrate was always monitored before and during the film growth and showed a similar RHEED pattern across the sample series.The evaporation rates are determined for each material from a calibrated thickness monitor (2% error), while the La 0.8 Sr 0.2 MnO 3 thickness was monitored in real time by following the RHEED intensity oscillations; for the buffer layer, it was not possible to determine the exact thickness exclusively on the basis of the RHEED oscillations since they are not observable at the early stages of deposition.The LaMnO 3 and La 0.45 Sr 0.55 MnO 3 thicknesses are hence estimated from the total growth time and the deposition rate, estimated by the RHEED oscillations appearing later in the deposition process.For this reason, we take into account an experimental error of 1 uc in the LaMnO 3 thickness.In Figure 1a−c, we show representative RHEED patterns for the TiO 2 -terminated SrTiO 3 substrate, LaMnO 3 , and La 0.8 Sr 0.2 MnO 3 films after growth, respectively, demonstrating epitaxial growth and good surface properties.The observation of RHEED intensity oscillations, as illustrated in Figure 1d for the 3 uc La 0.8 Sr 0.2 MnO 3 /LaMnO 3 film, indicates layer-by-layer growth; from atomic force microscopy (AFM) measurements, such as shown in Figure 1f, one observes the atomic steps from the substrate with a step height of ∼4 Å, i.e., of a single atomic unit cell, showing that the manganite films grow in single unit cell steps.After growth, the samples belonging to the LaMnO 3 series are annealed in air at 600 °C for 6 h to fully oxygenate the films.Henceforth, the samples are named with the thickness of the top La 0.8 Sr 0.2 MnO 3 layer in uc.The atomic force microscopy (AFM) measurements were performed in the tapping mode using a Bruker dimension icon 3100 instrument.
X-ray diffraction measurements were carried out in a Seifert fourcircle diffractometer equipped with a Cu anode X-ray tube and a linear array detector.The measurements consisted of θ−2θ scans along the (002) diffraction peak to determine the out-of-plane lattice parameter and reciprocal space mapping scans around the (103) diffraction peak to extract the in-plane lattice parameter.The transport properties were measured in a home-built four-probe setup in a van der Pauw configuration, 37,38 which uses spring contacts for a fast measurement setup; one drawback is that it leads to strong variations in the contact resistance, which we attribute to surface adsorbates due to the high vacuum or to surface residues from previous cleaning steps, making estimates of the absolute resistivity difficult.SQUID magnetometry measurements were performed using a 7 T Quantum Design MPMS3− 137 magnetic properties measurement system (MPMS) with the magnetic field applied along the film plane.X-ray absorption spectroscopy (XAS) measurements were carried out at the SIM and X-Treme beamlines at the Swiss Light Source (SLS) at the Paul Scherrer Institute (PSI), Switzerland, with the signal collected in the total electron yield (TEY) mode, which has a high surface sensitivity due to the limited mean free path of the electrons in solid materials on the order of 3 nm, or about 8 uc, and the exponential attenuation of the TEY signal with thickness.When measuring at SIM, before each magnetic XAS measurement, a saturation field of 1 kOe was applied to the sample and the measurements were performed under a 200 Oe magnetic field.At X-Treme, the saturation field was 6 T and the measurements were performed at remanence.We collected XAS spectra with circularly left and right polarized light, from which we obtained the unpolarized (average) spectra and the X-ray magnetic circular dichroism (XMCD) spectra (difference).The latter are normalized by dividing it by the L 3 peak intensity of the respective average XAS spectrum to give the dichroic signal as a percentage of the average XAS spectrum.Linear dichroism measurements were carried out by measuring the XAS spectra with the sample at a normal angle of incidence and at 45°to the X-ray beam to probe, respectively, the in-plane and out-of-plane orbital states; for one sample (5 uc La 0.8 Sr 0.2 MnO 3 /La 0.45 Sr 0.55 MnO 3 ), the X-ray linear dichroic (XLD) measurements were carried out using the X-ray photoemission electron microscope at the SIM beamline, where the light impinges the sample at a grazing angle of 16°.The different experimental conditions for these measurements (i.e., different monochromator grating and beam slit settings) result in a lower energy resolution of the 5 uc spectra.In order to compare spectra taken at different geometries, the out-of-plane XAS spectra are calculated using the expression for the angular dependence of the linear dichroism, I(θ) = I ab cos 2 θ + I c sin 2 θ, where θ is the angle of incidence of the light with respect to the sample surface and I ab and I c are the X-ray absorption parallel to the sample surface and perpendicular to it, respectively.

RESULTS AND DISCUSSION
To determine the strain state of the heterostructures, we carried out X-ray diffraction measurements on the two 10 uc La 0.8 Sr 0.2 MnO 3 films.The θ−2θ scans around the (002) SrTiO 3 peak are shown in Figure 2a,b for the LaMnO 3 and La 0.45 Sr 0.55 MnO 3 buffer layers, respectively.In both cases, broad peaks to the right of the (002) SrTiO 3 line are observed, corresponding to the pseudocubic (002) planes of the manganite films; in the case of the LaMnO 3 buffer layer, only one peak can be distinguished.By fitting the peak with three Gaussian components (Figure 2a), one for the SrTiO 3 peak together with a diffused scattering component and a third for the LaMnO 3 /La 0.8 Sr 0.2 MnO 3 bilayer, we obtain for the latter an outof-plane lattice parameter of 3.87 Å, which is below both LaMnO 3 and La 0.8 Sr 0.2 MnO 3 pseudocubic bulk values of a/√2 = 3.89 Å. 13 For the case of the film with the La 0.45 Sr 0.55 MnO 3 buffer layer, we observe the presence of clear Laue oscillations, indicative of high-quality interfaces and that the manganite peak is composed of two overlapping components (Figure 2b), which we ascribe to the La 0.8 Sr 0.2 MnO 3 and La 0.45 Sr 0.55 MnO 3 films.By fitting this peak to two Gaussian components together with the SrTiO 3 peak with a diffused scattering component, we obtain out-of-plane lattice parameters of 3.84 and 3.81 Å, which we assign to the La 0.8 Sr 0.2 MnO 3 and La 0.45 Sr 0.55 MnO 3 films, respectively.These values are considerably smaller than the respective pseudocubic bulk lattice parameters of 3.89 and 3.84 Å, 13,34,39 as expected for the tensile strain induced by the SrTiO 3 substrate.
Reciprocal space mapping (RSM) around the asymmetric (103) plane of SrTiO 3 is shown in Figure 2c,d for L a 0 .8 S r 0 . 2 M n O 3 / L a M n O 3 a n d L a 0 .8 S r 0 . 2 M n O 3 / La 0.45 Sr 0.55 MnO 3 , respectively.One finds that the Bragg reflection corresponding to the film bilayer lies at the same Q x value for both heterostructures, which directly shows that the manganite films grow coherently to the SrTiO 3 substrate and are fully strained.We expect likewise the thinner La 0.8 Sr 0.2 MnO 3 films to be also fully strained to the SrTiO 3 substrate.
The temperature dependence of the magnetization (M−T) of the ultrathin La 0.8 Sr 0.2 MnO 3 /LaMnO 3 films as a function of La 0.8 Sr 0.2 MnO 3 thickness is shown in Figure 3a.−47 The Curie temperature, T C = 145 K, is coincidentally very close to the Neél temperature of bulk LaMnO 3 , of 143 K. 13 While the origin of magnetism in LaMnO 3 is still under debate, recent surface-sensitive X-ray photoemission spectroscopy measurements indicate the presence of Sr and Ca in the LaMnO 3 film in both the as-grown and annealed states of up to a few %, suggesting that the magnetic moment in LaMnO 3 may be driven in large part by doping from divalent cations diffusing from the substrate. 48For the La 0.8 Sr 0.2 MnO 3 films, the results show that starting from 1 uc La 0.8 Sr 0.2 MnO 3 , the Curie temperature jumps to T C = 260 K (compared with 309 K for bulk La 0.8 Sr 0.2 MnO 3 and 305 K for thick La 0.8 Sr 0.2 MnO 3 / STO(001), slightly reduced from the bulk value due to epitaxial tensile strain); 13,17 with increasing La 0.8 Sr 0.2 MnO 3 thickness, T C is seen to oscillate around 250 K, as shown in inset I of Figure 3c, which could be a manifestation of finite-size effects. 3Notably, we find no independent contribution from the LaMnO 3 buffer layer in the M−T curves, indicating that the two layers are fully coupled magnetically.We also observe a steady increase in the saturation magnetization with increasing thickness, which we quantify from magnetic hysteresis curves carried out at 20 K, shown in Figure 3b.These data show that the La 0.8 Sr 0.2 MnO 3 films have small coercive fields, of ∼50 Oe, except for the 2 and 10 uc La 0.8 Sr 0.2 MnO 3 films, which exhibit coercive fields of ∼300 Oe.We attribute the larger coercive field to sample-to-sample growth-related variations.The larger coercivity implies stronger energy barriers to domain wall motion or to nucleation of reverse domains; given the high sensitivity of the coercive field to extrinsic factors, including local structural variations induced by the substrate morphology (such as degree of miscut, surface roughness, and point defects), some variation in the coercive field may be expected.By extrapolating from the high field region to zero field, we estimate the zero-field saturation magnetization as a function of thickness, shown in Figure 3c, to find that the saturation moment increases linearly with thickness, with a slope of 3.8 μ B /uc expected for bulk La 0.8 Sr 0.2 MnO 3 , shown as a red line.This result demonstrates that the insertion of a LaMnO 3 buffer layer results in La 0.8 Sr 0.2 MnO 3 films that are fully magnetically polarized down to 1 uc.The intercept at the origin gives the magnetization of LaMnO 3 , of 20.7 μ B /uc area, corresponding to an average magnetic moment of 1.7 μ B /Mn, in agreement with previous findings in the literature. 42,47Assuming the LaMnO 3 moment to be constant throughout the sample series, we can subtract its (constant) magnetic contribution to obtain the linear behavior shown in inset II of Figure 3c.For the 3 uc La 0.8 Sr 0.2 MnO 3 thickness, we expect a slightly larger magnetic moment due to the thicker LaMnO 3 layer (14 uc), and in inset II of Figure 3c, we show the magnetic moment after subtracting an additional moment of 3.4 μ B corresponding to the extra 2 uc LaMnO 3 .These results highlight the fact that the La 0.8 Sr 0.2 MnO 3 films develop a full magnetic moment starting from the first unit cell.In our discussion, we have supposed that no interdiffusion between the two layers occurs, i.e., that the two layers are welldefined.This is supported by our results: a systematic increase in the saturation magnetization as a function of the top La 0.8 Sr 0.2 MnO 3 thickness of 3.8 μ B /uc indicates that interdiffusion, if present, is not the major factor driving the observed phenomena.
The transport properties are listed in Figure 3d.A first remark is that the relative variation of the resistivity with La 0.8 Sr 0.2 MnO 3 film thickness does not follow the expected behavior, namely, a decrease with increasing film thickness.We attribute this largely to contact resistance introduced by the spring contacts in our measurement setup; coincidentally, the 2 and 10 uc La 0.8 Sr 0.2 MnO 3 films showing a larger coercive field also have a systematically larger resistivity, a property which is also very sensitive to local morphological and structural variations.However, while the presence of defects in the samples is expected to impact the resistivity and coercive field, we find that it does not strongly affect the saturation moments, as the M−H curves indicate.From the temperature dependence, we find that the 12 uc thick LaMnO 3 film is insulating, consistent with previous reports and as expected for bulk LaMnO 3 ; 13,33 importantly, starting from 1 uc thickness, the films exhibit a clear peak in the resistivity at a temperature near the magnetic critical temperature (inset I of Figure 3c), typical of the mixedvalence "colossal" magnetoresistance (CMR) manganites. 49he 2 and 4 uc La 0.8 Sr 0.2 MnO 3 films show a second transition to insulating behavior at low temperatures, which is often found in thin La 0.8 Sr 0.2 MnO 3 /SrTiO 3 (001) films and attributed to A-site disorder-induced charge localization effects. 50These results show that also the transport properties of the ultrathin La 0.8 Sr 0.2 MnO 3 films are similar to that of the bulk counterpart. 13he La 0.8 Sr 0.2 MnO 3 films are conducting and exhibit a metal-toinsulator transition starting from 1 uc, which demonstrates that the La 0.8 Sr 0.2 MnO 3 is in a metallic ferromagnetic state and has no electric dead layer as is the case with La 1−x Sr x MnO 3 films grown on SrTiO 3 , for example, and that the peak in resistivity agrees with T c obtained from the M−T curves, which is another characteristic of the mixed-valence manganites.These are important aspects for utilizing such ultrathin films for field effect devices or as polarizers in tunnel barriers.
The results for the case of La 0.8 Sr 0.2 MnO 3 /La 0.45 Sr 0.55 MnO 3 are markedly different.The magnetic hysteresis loops, presented in Figure 4a, show that the La 0.45 Sr 0.55 MnO 3 /SrTiO 3 (001) film exhibits no magnetic hysteresis, consistent with the expected antiferromagnetic state reported earlier. 34Also, the La 0.8 Sr 0.2 MnO 3 films up to 3 uc thickness show a negligible magnetic response to the applied magnetic field, while for thicknesses above 4 uc, magnetic hysteresis is present with coercive fields in the range from 50 to 100 Oe.The magnetic behavior is also confirmed by temperature-dependent magnetization measurements shown in Figure 4b for the 4, 5, 8, and 10 uc films, where the onset of magnetization in the system is observed at 250 K (with the 4 uc film showing an uptick in the magnetization at around 100 K; this behavior was reproduced in a second sample grown subsequently).Remarkably, the 10 uc La 0.8 Sr 0.2 MnO 3 film reaches a critical temperature of about 300 K, close to the bulk value.The saturation magnetic moment extracted from the M−H curves is presented in Figure 4c and shows an approximately linear increase in magnetization for thicknesses above 3 uc with a slope of about 3.8 μ B /uc.
The resistivity of the La 0.8 Sr 0.2 MnO 3 films grown on La 0.45 Sr 0.55 MnO 3 as a function of temperature is shown in Figure 4d.Since in this case the buffer layer brings a nonnegligible contribution to the total resistivity, the latter is calculated using the bilayer film thickness.The 0 uc sample shows a metal-to-insulator transition (MIT) at T MIT = 287 K, in agreement with a previous report. 34The reduced resistivity and increased MIT temperature (T MIT ) of the 1, 2, and 3 uc samples show that the La 0.8 Sr 0.2 MnO 3 layers have a lower resistivity compared to the buffer layer.The 4 uc La 0.8 Sr 0.2 MnO 3 / La 0.45 Sr 0.55 MnO 3 shows instead a drop in T MIT , as also shown in the figure inset (we confirmed this behavior from a second 4 uc La 0.8 Sr 0.2 MnO 3 /La 0.45 Sr 0.55 MnO 3 sample).A further increase in T MIT is observed for the 5 and 10 uc samples, saturating at ∼300 K.The critical temperature estimated from the peak resistivity is found to be significantly different from that estimated from the M−T curves, a discrepancy that is much larger than that for the La 0.8 Sr 0.2 MnO 3 /LaMnO 3 series.For example, for the 5 uc film, the critical temperature obtained from the magnetization curve is about 260 K, while the peak in resistivity remains close to that of 3 uc La 0.8 Sr 0.2 MnO 3 , of about 290 K. Differently from the previous case, where the LaMnO 3 buffer layer is insulating and T MIT is entirely determined by the La 0.8 Sr 0.2 MnO 3 layer itself, the La 0.45 Sr 0.55 MnO 3 buffer layer is conducting; hence, the transport properties are determined by the interplay of the two layers.Therefore, although a formal distinction of the two layers cannot be done from the resistivity data, we can draw some conclusions by combining the transport and magnetization data: up to 4 uc, La 0.8 Sr 0.2 MnO 3 couples antiferromagnetically to La 0.45 Sr 0.55 MnO 3 , which is inferred from the SQUID measurements and the systematic increase of T MIT within this thickness range.When La 0.8 Sr 0.2 MnO 3 becomes ferromagnetic, T MIT is given by a weighted average of the two contributions corresponding to the Curie (for La 0.8 Sr 0.2 MnO 3 ) and Neél (for La 0.45 Sr 0.55 MnO 3 ) temperatures.The fact that they do not agree suggests that at and above 5 uc thickness, the ferromagnetic component of the La 0.8 Sr 0.2 MnO 3 film is magnetically decoupled from the manganite layers underneath.An alternative explanation, given that one does not clearly distinguish two peaks in the resistivity, could be that the top La 0.8 Sr 0.2 MnO 3 films switch from an antiferromagnetic state at higher temperatures to the ferromagnetic ground state with decreasing temperatures.The kink at around 100 K observed for the 1−4 uc films is due to the SrTiO 3 cubic-to-tetragonal phase transition, which affects particularly strongly La 1−x Sr x MnO 3 films at near the 0.5 doping. 34,51e interpret the results for La 0.8 Sr 0.2 MnO 3 /La 0.45 Sr 0.55 MnO 3 as showing that the 1−3 uc La 0.8 Sr 0.2 MnO 3 films either are antiferromagnetic or couple antiferromagnetically to the La 0.45 Sr 0.55 MnO 3 buffer layer with the possible presence of antiferromagnetic domains in the La 0.45 Sr 0.55 MnO 3 buffer layer averaging out the net magnetization, while for larger thicknesses, the additional layer changes abruptly to a ferromagnetic state.Xray photoemission electron microscopy results show, in fact, that for the 1−3 uc La 0.8 Sr 0.2 MnO 3 film no ferromagnetic contrast over an antiferromagnetic multidomain state is present, consistent with the scenario that at small thicknesses, the thinner La 0.8 Sr 0.2 MnO 3 films order antiferromagnetically. 52The increase in the Neél temperature (corresponding to T MIT in these films) for 1−3 uc indicates an increase of the magnetic exchange interaction, while the large difference between T C and T MIT shows that the ferromagnetic layers are magnetically decoupled from the antiferromagnetic layer underneath.Starting from 4 uc thickness, the magnetic moment increases linearly with a slope of 3.8 μ B /Mn, indicated by the red line in Figure 4c, similar to the LaMnO 3 buffer layer case but shifted by 3 uc.We mention here the possibility for frustration at the interface, which was found to be present in La 1−x Sr x MnO 3 / La 0.6 Sr 0.4 FeO 3 multilayers, where La 0.6 Sr 0.4 FeO 3 is an antiferromagnetic system displaying a C-type canted antiferromagnetic spin state that leads to spin frustration at the La 1−x Sr x MnO 3 interface and to more complex spin configurations. 53In our situation, we expect La 0.45 Sr 0.55 MnO 3 to be in an A-type antiferromagnetic state, consisting of ferromagnetic spin planes coupled antiferromagnetically along the [001] direction.Such a magnetic spin configuration should not lead to spin frustration at the La 0.8 Sr 0.2 MnO 3 interface.
To better understand the magnetic and electronic properties of the La 0.8 Sr 0.2 MnO 3 films, we carried out X-ray magnetic spectroscopy measurements.In Figure 5a,b, we show the evolution of the nonpolarized XAS spectra with thickness for the films grown on the LaMnO 3 and La 0.45 Sr 0.55 MnO 3 buffer layers, respectively.For the case of the films grown on LaMnO 3 , we find no significant changes in the spectra across the sample series, which show features characteristic of fully oxidized La 0.8 Sr 0.2 MnO 3 . 54,55(The spectra corresponding to the 0 and 10 uc samples were collected at the X-Treme beamline and show a different peak amplitude, attributed to different background and measurement conditions between the two endstations; we aligned the energy scale between these two sets of measurements by matching the peak energy of the 10 uc film to that of the 5 uc).Although one may expect a small shift in the peak energy position when going from LaMnO 3 to La 0.8 Sr 0.2 MnO 3 , the shift is relatively small, 56 compounded by the fact that the LaMnO 3 film is itself slightly doped.The corresponding normalized XMCD spectra are shown in Figure 5c.Similar to XAS, we find no significant changes in the shape and amplitude of the XMCD spectra; the 0 and 10 uc samples show a reduced and an enhanced XMCD signal, respectively.These results confirm that the magnetic moment originates from the Mn cations and are in agreement with the bulk magnetometry results showing full magnetic moment of the La 0.8 Sr 0.2 MnO 3 film: since the spectra are collected in TEY, which has a sensitivity of around 3−5 nm due to the limited electron mean free path, and is therefore representative of the magnetic state of the surface layers, 57 a constant XMCD signal with thickness confirms that each added layer is magnetic.Interestingly, since the amplitude of the LaMnO 3 buffer layer XMCD is comparable to that of the other samples in the series and the XMCD amplitudes are normalized to the XAS spectra, our results suggest that the magnetic moment in LaMnO 3 arises from the surface layer.The La 0.45 Sr 0.55 MnO 3 sample series was measured at the SIM beamline, with the exception of the 5 uc, measured at X- Treme.The shape of the main peak at 642 eV is an indication that the films are stoichiometric and fully oxidized, with Mn present in a mixed 3+/4+ valence state. 56Figure 5d shows the corresponding XMCD spectra.Consistent with the magnetometry results, the 0, 1, and 2 uc samples show no XMCD signal (the residual signal for 0 uc results from a slight difference in amplitude for the circular right and left light that could not be corrected in the data and does not have the expected XMCD energy variation), while the 5 uc sample displays a strong magnetic dichroic response, but with an amplitude that is lower than the case of the LaMnO 3 series.These results support the conclusion that the magnetic signal arises from the top two unit cells and is slightly reduced by the zero XMCD signal contribution from the antiferromagnetic La 0.8 Sr 0.2 MnO 3 / La 0.45 Sr 0.55 MnO 3 films underneath.
Magnetic exchange in these materials is generally understood in terms of orbital occupancy, according to the Goodenough− Kanamori rules. 12,58In particular, doping and strain play fundamental roles in determining which of the spin exchange mechanisms, double exchange or superexchange, dominates the magnetic interaction since the orbital structure depends strongly on those parameters.In order to probe the orbital character of the La 0.8 Sr 0.2 MnO 3 /La 0.45 Sr 0.55 MnO 3 heterostructures, we carried out X-ray absorption measurements with linearly polarized light, which probes the available density of states along the electric field direction. 59The XAS spectra measured at 300 K, above the Curie temperature to avoid the magnetic contribution, are shown in Figure 6a,d for the O K-edge and Mn L-edge, respectively.
The O K-edge shows the general shape typical for La 0.8 Sr 0.2 MnO 3 , with pre-edge peaks (A, B) associated with O 2p orbitals hybridized with Mn 3d and higher energy peaks, C, D, associated with O 2p orbitals hybridized with La and Sr orbitals, respectively. 56Of interest here are the pre-edge peaks, in the range 527−533 eV, which are argued to reflect more directly the unoccupied density of states of the Mn 3d states, since the excitation occurs at the oxygen site. 60Two different assignments of these peaks to the Mn orbitals states have been presented in the literature, in one case where peak A is assigned to the partially filled e g ↑ band and peak B to transitions into empty t 2g ↓ and e g ↓ states; 56,61,62 another assignment ascribes peak A to the partially filled e g ↑ band and empty t 2g ↓ and peak B to transitions into e g ↓ states. 63Our results are not able to address this controversy; however, we can extract the following conclusions from our data.(i) One finds a shift to higher energies of the O pre-edge features associated with the e g ↑ band when going from 0, 1, and 2 uc La 0.45 Sr 0.55 MnO 3 to 5 and 10 uc La 0.8 Sr 0.2 MnO 3 (Figure 6b); this is expected based on band filling, since with decreasing hole doping the electron occupation of the e g ↑ band increases, resulting in a decrease in the unoccupied density of states.(ii) One finds that the leading edge of the X-ray absorption spectra occurs at lower energies for the in-plane polarized light, while the difference is significantly reduced for the 5 and 10 uc films; this can be more easily seen as a shift to higher energies of the spectral weight of the linear dichroic signal at the e g ↑ band, Figure 6c.This behavior indicates a reduction in the asymmetry of the occupancy of in-plane and out-of-plane orbitals at larger La 0.8 Sr 0.2 MnO 3 thicknesses, consistent with what is expected for the ferromagnetic state. 64Furthermore, the role of strain is that of shifting the relative energy position of the two e g ↑ orbitals, z 2 and x 2 − y 2 , with compressive (negative) strain favoring the x 2 − y 2 orbitals. 65We note that, given that both layers have the same in-plane lattice parameter and given that the lattice parameter of La 1−x Sr x MnO 3 changes with Sr doping, the La 0.45 Sr 0.55 MnO 3 buffer layer and the La 0.8 Sr 0.2 MnO 3 top layer are subjected to different inplane strains.Taking the pseudocubic lattice parameters for La 0.45 Sr 0.55 MnO 3 and La 0.8 Sr 0.2 MnO 3 as 3.84 and 3.89 Å, respectively, 13,34,39 one obtains in-plane strains of −1.69 and −0.39%, respectively.Thus, our results agree with the expectations and are consistent with the magnetic behavior, i.e., an antiferromagnetic in-plane metallic A-type state for La 0.45 Sr 0.55 MnO 3 and a ferromagnetic isotropic state for La 0.8 Sr 0.2 MnO 3 .(iii) In terms of the thickness evolution, one finds that up to 2 uc La 0.8 Sr 0.2 MnO 3 the XLD spectra in Figure 6e remain essentially unchanged (although in these cases, a significant signal contribution from the underlying La 0.45 Sr 0.55 MnO 3 layer should be present), while for 5 uc, which is expected to have a strongly reduced La 0.45 Sr 0.55 MnO 3 signal contribution due to the limited electron escape depth, a strong modification in the XAS spectra is observed, starting to resemble the 10 uc La 0.8 Sr 0.2 MnO 3 spectra.(iv) The integral of the linear dichroism over the Mn L 2,3 edge, Figure 6e, gives the preferred hole orbital occupation according to the linear dichroism sum rule D L = ∫ Ld 2 +Ld 3 (μ ab − μ c ) dE/∫ Ld 2 +Ld 3 (2μ ab + μ c )dE ∝ ∑⟨l z 2 − 2⟩ i , where μ ab and μ c are the X-ray absorption along the in-plane and out-ofplane directions, respectively, l z is the quantum orbital number, with l z = 0 for the 3d 3z 2 −r 2 orbital and l z = 2 for 3d x 2 −y 2 . 66A positive value for D L indicates a preferential hole occupation of the x 2 − y 2 orbitals.Based on this sum rule, one finds, qualitatively, that for 0−2 uc La 0.8 Sr 0.2 MnO 3 the in-plane x 2 − y 2 orbitals are preferentially occupied by hole carriers, 61,67 while for 5 and 10 uc, there is a change in the orbital character to a predominant hole occupation of the out-of-plane orbitals.The changes in the orbital character are in general qualitative agreement with the O pre-edge spectral features.
We also carried out magnetic linear dichroism (XMLD) measurements for 0, 1, 2, and 10 uc La 0.8 Sr 0.2 MnO 3 / La 0.45 Sr 0.55 MnO 3 at 20 K, as shown in Figure 7.The XMLD signal is normalized to the XAS peak intensity, and these measurements are taken with a large beam spot (of about 1 mm in lateral size), averaging over a large area of the sample, i.e., it is an average over the antiferromagnetic multidomain state of the sample.Assuming that for all samples there is a similar antiferromagnetic domain distribution, these data tell us that the amplitude of the XMLD signal remains approximately constant with increasing film thickness, which indicates that each added layer contributes similarly to the XMLD signal (otherwise, the relative magnetic contribution to the total signal would drop).These data further support our conclusion that the top La 0.8 Sr 0.2 MnO 3 layer is in an antiferromagnetic state at lower thicknesses, as discussed above.
From these observations, in combination with the magnetic behavior, we conclude that La 0.8 Sr 0.2 MnO 3 films up to 3 uc thickness have a similar electronic and magnetic structure as the La 0.45 Sr 0.55 MnO 3 buffer layer, characterized by a preferential hole occupation of the x 2 − y 2 orbitals, large in-plane conductivity, and an A-type antiferromagnetic ordering.At larger thicknesses, a transition to a preferential occupation of out-of-plane orbitals occurs, where a spectrum similar to that reported for the fully polarized ferromagnetic state is observed, 68 even if an equal population of both in-plane and out-of-plane orbitals were expected. 69he two different buffer layers, therefore, lead to two distinct magnetic behaviors for the top La 0.8 Sr 0.2 MnO 3 films.In the case of LaMnO 3 , the La 0.8 Sr 0.2 MnO 3 films are fully magnetically polarized starting from 1 uc, with a magnetic critical temperature that is much higher than that of the single buffer layer, indicating the presence of a strong magnetic coupling dominated by the top La 0.8 Sr 0.2 MnO 3 .Since the LaMnO 3 film is insulating, we attribute the magnetic coupling to be dominated by double exchange at the interface between the two systems.In the case of the La 0.45 Sr 0.55 MnO 3 buffer layer, we find that the La 0.8 Sr 0.2 MnO 3 film adopts the metallic antiferromagnetic state of the buffer layer up to a thickness of 3−4 uc and a sudden transition to a ferromagnetic state above this thickness, manifested in the linear dichroic data by a sudden modification of the electronic structure from 2 to 5 uc films.We explain the interfacial antiferromagnetic metallic state as driven by extended delocalized charge transfer between La 0.45 Sr 0.55 MnO 3 and La 0.8 Sr 0.2 MnO 3 , which operates up to three unit cells beyond the interface.At larger thicknesses, the La 0.8 Sr 0.2 MnO 3 adopts its ferromagnetic ground state with a full bulk moment and critical temperature independent of the underlying manganite film.In both instances, the La 0.8 Sr 0.2 MnO 3 films are fully magnetically polarized, starting from the single unit cell.One possible common mechanism driving the magnetic behavior of the top La 0.8 Sr 0.2 MnO 3 films toward bulk behavior may be the onset of oxygen octahedral rotations promoted by the LaMnO 3 buffer layer and the first 3−4 unit cells La 0.8 Sr 0.2 MnO 3 / La 0.45 Sr 0.55 MnO 3 , which agrees with the length scale for bond angle relaxation of the oxygen octahedral tilt of 4 unit cells found for LaMnO 3 /SrTiO 3 superlattices. 70ur results are highly relevant for device applications since they provide an approach to using the same material platform to create La 0.8 Sr 0.2 MnO 3 films that are fully magnetically polarized down to 1 uc either on conducting buffer layers (La 0.45 Sr 0.55 MnO 3 ) or on magnetic but insulating layers (LaMnO 3 ).We anticipate as well that the results found here may be applicable to the wider doping range where La 1−x Sr x MnO 3 is in a ferromagnetic state, including optimally doped La 1−x Sr x MnO 3 , and to the wider manganite perovskite family.

CONCLUSIONS
In conclusion, we have studied the evolution of the magnetic moment of ultrathin La 0.8 Sr 0.2 MnO 3 films grown on insulating LaMnO 3 and conducting La 0.45 Sr 0.55 MnO 3 buffer layers.Although both systems are nominally antiferromagnetic, we find a significant magnetic moment in the LaMnO 3 buffer layer, a result similar to that reported previously in the literature.For La 0.8 Sr 0.2 MnO 3 /LaMnO 3 , a metallic state and bulk-like magnetic moment are observed for La 0.8 Sr 0.2 MnO 3 thicknesses down to 1 uc.The two layers are magnetically coupled, but the magnetic properties are dominated by the top La 0.8 Sr 0.2 MnO 3 film, including bulk-like magnetic moments and high critical temperatures that do not correspond to those of the bare LaMnO 3 film.For the case of La 0.8 Sr 0.2 MnO 3 /La 0.45 Sr 0.55 MnO 3 , we observe an antiferromagnetic ground state up to 3 uc, above which a ferromagnetic ground state for the subsequent La 0.8 Sr 0.2 MnO 3 layers emerges.The X-ray absorption spectroscopy results show a sudden change in the electronic structure from 2 to 5 uc, consistent with the observed change in the magnetic properties.Our results show that the properties of ultrathin La 0.8 Sr 0.2 MnO 3 films can be tailored by a suitable choice of buffer layers to yield bulk-like magnetic polarization and high critical temperatures, down to the unit cell thickness range, paving the way to fully exploring the unique electronic properties of this class of strongly correlated oxide materials.

Data Availability Statement
The data underlying this study are openly available at the PSI Public Data Repository database at doi.psi.ch/detail/10.16907/b927e658-d5b2-4892-aebf-6d831956939d.

Figure 1 .
Figure 1.(a−c) RHEED patterns of the TiO 2 -terminated SrTiO 3 (001) substrate, the LaMnO 3 buffer layer, and the 3 uc La 0.8 Sr 0.2 MnO 3 /LaMnO 3 film, respectively, along the ⟨100⟩ azimuth.The energy of the incident electron beam was set to 15 keV.(d) RHEED intensity oscillations as a function of time for 3 uc La 0.8 Sr 0.2 MnO 3 deposited on LaMnO 3 .(e) Schematic of the sample structure.(f) AFM measurement of the 10 uc of La 0.8 Sr 0.2 MnO 3 / LaMnO 3 sample, showing a flat surface with single unit cell step terraces (inset).

Figure 3 .
Figure 3. (a) Field-cooled magnetization variation with temperature of the La 0.8 Sr 0.2 MnO 3 /LaMnO 3 films taken during heating (applied magnetic field of 1 kOe).(b) Magnetic hysteresis curves at 20 K. Inset: Zoomed-in M−H curves highlighting the low field magnetic behavior.(c) Zero-field saturation magnetization per unit cell area versus La 0.8 Sr 0.2 MnO 3 film thickness (symbols) extrapolated from the M-H loops; the red line has a slope of 3.8 μ B /uc. Inset (I) shows the variation of the critical temperature (SQUID) and of the peak in resistivity as a function of La 0.8 Sr 0.2 MnO 3 film thickness (dashed lines are guides to the eye); inset (II) shows the magnetic moment (μ) variation without the LaMnO 3 moment contribution.The error bars in the magnetization are related to the experimental uncertainty due to the LaMnO 3 thickness.(d) Resistivity versus temperature for the La 0.8 Sr 0.2 MnO 3 / LaMnO 3 sample series.

Figure 4 .
Figure 4. (a) Magnetization hysteresis curves of the La 0.8 Sr 0.2 MnO 3 /La 0.45 Sr 0.55 MnO 3 sample series (20 K) and (b) magnetization temperature dependence of the 0, 4, 5, 8, and 10 uc La 0.8 Sr 0.2 MnO 3 films (1 kOe).(c) Zero-field saturation moments extrapolated from the M−H loops.The red line has a slope of 3.8 μ B /uc crossing the abscissa at 3 uc.Inset: Schematic representation of the magnetic spin configuration of the La 0.8 Sr 0.2 MnO 3 film on the La 0.45 Sr 0.55 MnO 3 buffer layer.(d) Transport measurements showing the evolution of the temperature-dependent resistivity with thickness.The metal-to-insulator transition variation with thickness, identified as T C , is shown in the figure inset together with the Curie temperature, obtained as a linear extrapolation to zero from the M−T curves.The diamond symbols in (c) and (d) are from a second 4 uc La 0.8 Sr 0.2 MnO 3 /La 0.45 Sr 0.55 MnO 3 sample grown subsequently.

Figure 5 .
Figure 5. XAS spectra collected on La 0.8 Sr 0.2 MnO 3 films grown on (a) the LaMnO 3 and (b) the La 0.45 Sr 0.55 MnO 3 buffer layer.The corresponding XMCD spectra, normalized to the respective XAS peak value, are shown in (c) and (d).All spectra are shifted vertically for clarity of display.

Figure 6 .
Figure 6.(a) Room-temperature XAS spectra at the O K-edge obtained with linearly polarized light for La 0.8 Sr 0.2 MnO 3 /La 0.45 Sr 0.55 MnO 3 .(b) Detail of the pre-edge feature of the spectra (out-of-plane light polarization) highlighting the transition associated with the O 2p-Mn 3d hybridized states.The derivatives of the 1 and 10 uc spectra are also shown to highlight the peak position.(c) O K-edge XLD spectra of the La 0.8 Sr 0.2 MnO 3 films.(d) XAS spectra at the Mn L-edge collected with vertical (I ab ) and horizontal (I c ) light polarization.Inset shows a plot of all spectra at the L 3 -edge.(e) Mn Ledge XLD spectra.Spectra are shifted vertically for clarity of display.