Top-layer engineering reshapes charge transfer at polar oxide interfaces

www.advmat.de 2203071 (1 of 9) © 2022 The Authors. Advanced Materials published by Wiley-VCH GmbH Top-Layer Engineering Reshapes Charge Transfer at Polar Oxide Interfaces Gabriele De Luca,* Jonathan Spring, Moloud Kaviani, Simon Jöhr, Marco Campanini, Anna Zakharova, Charles Guillemard, Javier Herrero-Martin, Rolf Erni, Cinthia Piamonteze, Marta D. Rossell, Ulrich Aschauer, and Marta Gibert* G. De Luca,[+] J. Spring, S. Jöhr, M. Gibert[++] Department of Physics University of Zurich Winterthurerstrasse 190, Zurich 8057, Switzerland E-mail: [email protected]; [email protected] DOI: 10.1002/adma.202203071 the emergence of a 2D electron liquid at the interface between LaAlO3 (LAO) and SrTiO3 (STO),[1] but other investigations involving transition metal cations with partially filled d-bands revealed that interfacial ferromagnetism,[2] exchange bias,[3] and orbital reconstructions[4,5] can be targeted with proper material design.[6] Interfacial charge-transfer phenomena, however, are also associated with detrimental effects on the desired functionalities, especially in the ultrathin limit.[7,8] A favorable strategy to control the extent of charge transfer consists in inserting a specific interlayer that modulates the electrostatic screening at the interface.[9] Unfortunately, this approach cannot be used if one would like to retain certain properties of the original interface (for instance, octahedral connectivity,[10,11] or substrate-driven field effect[12,13]). Therefore, alternative methods to influence the electronic configuration of a buried oxide interface are very intriguing.[14,15] In this work, we demonstrate that the heterointerface between a polar ferromagnetic insulating double perovskite thin film (La2NiMnO6, LNMO) and a non-polar single perovskite substrate (STO) is a prototypical example of an interfacial charge transfer triggered by a polarity mismatch.[16] The Charge-transfer phenomena at heterointerfaces are a promising pathway to engineer functionalities absent in bulk materials but can also lead to degraded properties in ultrathin films. Mitigating such undesired effects with an interlayer reshapes the interface architecture, restricting its operability. Therefore, developing less-invasive methods to control charge transfer will be beneficial. Here, an appropriate top-interface design allows for remote manipulation of the charge configuration of the buried interface and concurrent restoration of the ferromagnetic trait of the whole film. Doubleperovskite insulating ferromagnetic La2NiMnO6 (LNMO) thin films grown on perovskite oxide substrates are investigated as a model system. An oxygenvacancy-assisted electronic reconstruction takes place initially at the LNMO polar interfaces. As a result, the magnetic properties of 2–5 unit cell LNMO films are affected beyond dimensionality effects. The introduction of a top electron-acceptor layer redistributes the electron excess and restores the ferromagnetic properties of the ultrathin LNMO films. Such a strategy can be extended to other interfaces and provides an advanced approach to fine-tune the electronic features of complex multilayered heterostructures. ReseaRch aRticle 1. Introduction In the last two decades, a growing interest in charge-transfer phenomena in oxide heterostructures resulted in the discovery of many interfacial properties that are not exhibited by the bulk counterparts. The most famous example concerns © 2022 The Authors. Advanced Materials published by Wiley-VCH GmbH. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. The ORCID identification number(s) for the author(s) of this article can be found under https://doi.org/10.1002/adma.202203071. M. Kaviani, U. Aschauer Department of Chemistry Biochemistry and Pharmaceutical Sciences University of Bern Freiestrasse 3, Bern 3012, Switzerland M. Campanini, R. Erni, M. D. Rossell Electron Microscopy Center Empa Swiss Federal Laboratories for Materials Science and Technology Überlandstrasse 129, Dübendorf 8600, Switzerland A. Zakharova, C. Piamonteze Swiss Light Source Paul Scherrer Institut Villigen 5232, Switzerland C. Guillemard, J. Herrero-Martin ALBA Synchrotron Light Source Carrer de la Llum 2-26, Cerdanyola del Vallès 08290, Spain [+]Present address: Catalan Institute of Nanoscience and Nanotechnology (ICN2), Campus UAB, 08193, Bellaterra, Barcelona, Spain [++]Present address: Institute of Solid State Physics, TU Wien, Wiedner Hauptstrasse 8-10, 1040, Vienna, Austria Adv. Mater. 2022, 34, 2203071 15214095, 2022, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202203071 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [26/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License www.advmat.dewww.advancedsciencenews.com 2203071 (2 of 9) © 2022 The Authors. Advanced Materials published by Wiley-VCH GmbH resulting electronic reconstructions strongly affect the magnetic properties of the heterostructure, adding to dimensionality effects. To restore the optimal electronic configuration of the buried interface, we design a top-interface engineering approach through a LaNiO3 (LNO) overlayer. The presence of LNO is fundamental in displacing the interfacial charge and allows reestablishing the ferromagnetism in LNMO. This method can be extended to other systems and certifies that top-layer engineering can be successfully exploited to remotely control the electronic configuration of buried complex oxide interfaces. 2. Results and Discussion 2.1. Magnetic Properties of La2NiMnO6//SrTiO3 (LNMO//STO) Heterostructures LNMO is a ferromagnetic insulator with a bulk Curie temperature (TC) around 270 K, resulting from strong positive superexchange between the long-range-ordered Ni2+ (3d8) and Mn4+ (3d3) cations in a rock-salt-type structure.[17–19] To evaluate dimensionality and interface effects on LNMO magnetic and electronic properties, we grew a set of films of different thicknesses, ranging from 90 to 2 pseudocubic unit cell (pc uc), on prototypical STO substrates (tensile strain =+0.7%). Structural characterization indicates epitaxial growth and high crystalline and surface quality in all the films (Figure S1, Supporting Information).[19] As shown in Figure1a, films thicker than 30 uc (≈12nm) exhibit bulk-like ferromagnetic properties, that is, TC≈ 270 K and saturation magnetization larger than 4 µB/formula units.[19] However, the Curie temperature and the magnetization decrease as films are grown thinner. Still, 15 and 5 uc-thick films display TC≈ 250 K and TC≈ 165 K, respectively. This magnetic trend is further confirmed by X-ray magnetic circular dichroism (XMCD) measurements. The reduction of both Mn and Ni L3,2 integrals in normalized XMCD spectra is indicative of a decrease in the magnetic moment of both the Mn and Ni sublattices in the ultrathin limit (L3-edges shown in Figure1b-c, the full L3,2-edges displayed in Figure S2, Adv. Mater. 2022, 34, 2203071 Figure 1. a) Magnetization versus temperature measured in a magnetic field of 0.5 T for various LNMO film thicknesses. b) Mn L3-edges XMCD spectra measured at 20 K and 5 T. c) Same as (b) for Ni. d) Integral of the Ni XMCD L3-edge as a function of the sample temperature for selected film thicknesses. Scaled data from a SQUID is shown for comparison (dashed and dotted lines). In the right panel, the behavior of the ultrathin 2 uc film is highlighted. Here, the line is a guide to the eye. e) XMCD asymmetry as a function of the magnetic field calculated at both Mn and Ni L3-edges for the 90 uc (top, 20 K), 3 uc (middle, 20 K), and 2 uc (bottom, 5 K) LNMO films. The lines are guides to the eye. 15214095, 2022, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202203071 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [26/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License www.advmat.dewww.advancedsciencenews.com 2203071 (3 of 9) © 2022 The Authors. Advanced Materials published by Wiley-VCH GmbH Supporting Information). In Figure1d, we plot the evolution of the Ni XMCD L3-integral as a function of the temperature as an alternative way to determine the Curie temperature. The elemental sensitivity of XMCD compared to SQUID magnetometry allows films as thin as 2 uc (≈0.8nm) to be characterized. We find that 2 uc films have TC≈ 10 K. To gain insight into the origin of this thickness-dependent degradation of the magnetic properties, it is instructive to observe the difference between the Mn and Ni XMCD asymmetry measured as a function of the external magnetic field for different film thicknesses, as displayed in Figure1e. For a bulklike 90-uc-thick film, both the Mn and Ni asymmetry saturate at similar values of the magnetic field. For a 3-uc-thick film, we still observe a similar behavior between Mn and Ni despite asymmetries saturating at larger fields. For the 2-uc-thick film, the Ni sublattice behavior is still reminiscent of a ferromagnetic hysteresis whereas the Mn sublattice is severely affected by the thickness reduction, nearly losing its ferromagnetic character. 2.2. Charge Transfer Induced by Interfacial Polarity Mismatch Since the electronic configuration of Ni and Mn is key to the ferromagnetic behavior of the LNMO system,[17] we investigated the valence state of the cations using X-ray absorption spectroscopy (XAS). In Figure2a, we show the L3,2-edges of Mn as a function of the LNMO thickness. The XAS data of the 90 uc sample is consistent with previous observations on bulk specimens and certifies the valence state of Mn to be prevalently 4+.[20] As the film thickness is reduced, a pronounced shoulder in the low-energy range and an increased spectral weight around 641eV emerge. These features can be associated with Adv. Mater. 2022, 34, 2203071 Figure 2. a) XAS of Mn L3,2-edges for different LNMO thicknesses (top). Data is collected at 20 K. Reference XAS spectra for various electronic configurations of Mn[22] (bottom). b) Same as (a) for Ni L2-edge.[23] The Ni L3-edge overlaps with the La M4-edge and it is not shown. In the inset, a zoom on the Ni L2-edge maxima highlights the progressive shift to lower energies for decreasing LNMO thickness. c) Sketch of interfacial stacking sequence of LNMO//STO. The nominal ionic charge of each layer is indicated on the left. d) pDOS for the Ni0.5Mn0.5O2 layer adjacent to the LNMO//STO interface. Mn states at the Fermi level indicate the presence of charge transfer. e) pDOS for the interfacial Ni0.5Mn0.5O2 layer in presence of oxygen vacancies. In this case, both majority spin Mn and minority spin Ni states are populated by the excess electrons. The overlap between Mn and Ni states is due to orbital hybridization. 15214095, 2022, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202203071 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [26/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License www.advmat.dewww.advancedsciencenews.com 2203071 (4 of 9) © 2022 The Authors. Advanced Materials published by Wiley-VCH GmbH an increase of Mn3+ content located at the interface between LNMO and STO. However, we cannot exclude the presence of additional Mn3+ states present also in the top surface layer, as we previously detected in 30-uc-thick LNMO//STO heterostructures.[19] We also observe a small alteration of the Ni valence state (Figure2b) to which we will come back. From previous structural characterizations, we know that our relatively thick LNMO films grow with a prevalent rocksalt ordering of Ni2+ and Mn4+ cations and display bulk-like TC and saturation magnetization.[19] Therefore, we can exclude sub-optimal growth conditions. It is known that in perovskite oxide films the formation energy of bulk oxygen vacancies is reduced with increasing tensile strain.[21] A systematic XAS study on 5 uc-thick LNMO films grown on a variety of substrates (imposing epitaxial strains ranging from −2.1% to +1.8%[19]) reveals that the LNMO film with the largest Mn3+ content is the one grown on STO, followed by the one grown on LSAT (Figure S3, Supporting Information). As none of these two substrates exerts the largest tensile strain on the films, the strain-triggered oxygen-vacancy-formation scenario cannot be the dominant mechanism involved. Instead, we should recall that the ABO3 perovskite crystal structure can be thought of as a sequence of AO and BO2 planes that carry a nominal ionic charge depending on the specific A and B cations. This applies to the LNMO//STO interface which can be pictured as a sequence of uncharged (SrO)0 and (TiO2)0 layers followed by charged (LaO)+ and (Ni0.5Mn0.5O2)− stacks along the [001]pc orientation (Figure2c). Hence, the polar discontinuity at the interface builds up a diverging potential, generally referred to as “polar catastrophe” that can be relieved by an electronic reconstruction.[16] To investigate the implications of this, we performed density functional theory (DFT) calculations. Layer-resolved projected density of states (pDOS) for the stoichiometric LNMO//STO heterostructure without oxygen vacancies reveals the appearance of Mn electronic states located at the Fermi level in the Ni0.5Mn0.5O2 layer adjacent to the TiO2-interface layer as a mechanism to counteract the polar discontinuity (Figure 2d). Interestingly, this electron redistribution is not accompanied by a concomitant change in the Ni electronic states. Equivalent calculations for an LNMO//LaGaO3 (LGO) heterostructure, which is characterized by a similar tensile strain (=+0.4%) but without polar discontinuity, show no trace of extra states (Figure S4, Supporting Information). These results support the emergence of an electronic reconstruction at the LNMO// STO interface with the interfacial electrons preferentially localized at the Mn sites, similar to what was already observed in LaMnO3//STO.[24,25] From the DFT calculations, the built-in electric field within the LNMO film is estimated to be about 0.118V Å−1. Using the computed LNMO bandgap (≈1.66 eV) to estimate the critical LNMO thickness for the polar-discontinuity-induced electronic reconstruction, we obtain ≈3.65 uc (≈14 Å). Experimentally, however, we observe an electron redistribution in films as thin as 2 uc (Figure 2a,b), suggesting that the ionic picture alone is not sufficient to explain the electronic configuration of our ultrathin LNMO films. It has been argued that a discontinuity in the interface polarity can also be a driving force for the formation of oxygen vacancies.[26,27] Note that the origin of these interfacial oxygen vacancies is not epitaxial strain[21] but rather the polar discontinuity itself. To contemplate this possibility, we performed additional DFT calculations introducing oxygen vacancies in our heterostructure (Figure S5, Supporting Information). We find it is most favorable for the system to form oxygen vacancies when these are located at the film surface, followed by positions in the Ni0.5Mn0.5O2 interfacial slab. Breaking substrate TiO bonds, instead, leads to higher formation energies. In this scenario, due to extra electron doping from the oxygen vacancies and a defect-state-induced band gap reduction, an LNMO thickness smaller than 3.65 uc is enough to trigger the electronic reconstruction, in line with our XAS observations. These calculations further reveal that a few of the oxygen-vacancies-induced doping electrons are now also populating the minority spin Ni states in the Ni0.5Mn0.5O2 layer adjacent to the substrate, besides Mn (Figure2e and Figure S5b, Supporting Information). This indicates the appearance of Ni(2−x)+ states at the LNMO//STO interface modeled with oxygen vacancies, in contrast with the simple ionic picture calculations previously discussed. Experimentally, when decreasing the LNMO thickness on STO substrates, we observe a shift to lower energies of the Ni L2-edge spectral weight (Figure2b) suggesting a small reduction of the Ni valence state at the LNMO interfacial layers. Thus, both the calculations and the experimental observations indicate that a complex polarity-induced charge transfer mediated by oxygen vacancies is taking place within the LNMO films to compensate their polar discontinuity with the substrate. This scenario intuitively explains the presence of Mn3+ also observed at the (less) polar LNMO//LSAT interface (Figure S3, Supporting Information).[19,28] Notice that all other investigated substrates constitute nonpolar interfaces with LNMO and therefore in these heterostructures we always observe a reduced amount of Mn3+ compared to the STO and LSAT ones. 2.3. Atomically Resolved Cation and Electron Distribution To further verify our observations, we performed energy-dispersive X-ray (EDX) and electron energy-loss spectroscopy (EELS) analyses across the LNMO//STO heterostructures in a scanning transmission electron microscope (STEM). By imaging the films along the [ 110]pc LNMO zone axis we get insight into the NiMn cationic order.[19,29] The high-angle annular dark-field (HAADF) STEM image of a representative 13 uc LNMO//STO film is presented in Figure3a together with the combined EDX elemental maps obtained for Mn, Ni, and Ti (Figure3b). The individual elemental maps are shown in Figure S6, Supporting Information. We observe epitaxial coherent growth with absence of defects such as dislocations or precipitates through the entire surveyed area. From the EDX map, we deduce that the first layer is partially intermixed with Ti, as often realized at the interface between a polar transition metal oxide and STO.[30,31] Note that this tendency of the transition metal cation and Ti to swap is generally not enough to screen the interfacial polarity.[32] Away from the interface, we see that the film grows in a perfectly ordered manner,[19] as evidenced by the checkerboard patterns observed in the EDX maps. Closer to the interface, we can appreciate the existence of a transition region of Adv. Mater. 2022, 34, 2203071 15214095, 2022, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202203071 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [26/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License www.advmat.dewww.advancedsciencenews.com 2203071 (5 of 9) © 2022 The Authors. Advanced Materials published by Wiley-VCH GmbH around 5–6 uc with only partial rock-salt ordering of the LNMO film. DFT calculations considering various NiO6 and MnO6 octahedra double-perovskite configurations, namely rock salt, layered and columnar, show that rock-salt ordering is the most stable configuration for LNMO. However, the presence of an additional electron, such as the one transferred to the interface due to the polar discontinuity, turns the columnar ordering into the most favorable structure (Figure S7, Supporting Information). Hence, in the proximity of the Mn3+ ions located at the LNMO//STO heterointerface an increased tendency to B/B′-site disorder is expected, in agreement with the EDX elemental maps shown in Figure 3b and Figure S6, Supporting Information. Further, our STEM-EELS data confirm the formation of oxygen vacancies both in the Ni0.5Mn0.5O2 interfacial layers and at the film surface. In Figure3c, we show the O K and Mn L3,2 edges acquired across the LNMO film thickness at the layer positions indicated with the colored dots in Figure3a. The EEL spectra acquired at the middle Ni0.5Mn0.5O2 layers (layers 3 to 12 from the interface) certify the valence state of Mn to be prevalently 4+. However, at the Ni0.5Mn0.5O2 interfacial layers, the Mn L3 edge shifts to lower energies (from ≈642 to ≈641eV), and the O Ka pre-edge peak shifts to higher energies (from ≈528 to ≈529eV). An even larger shift of the Mn L3 edge of ≈2eV is observed at the film surface, while the O Ka pre-edge peak almost disappears. These observations are associated with an increase of Mn3+ content and oxygen vacancies located at the LNMO//STO interface and at the LNMO film surface,[33,34] which is in excellent agreement with the DFT calculations. We can also use the EELS data to quantify the local Mn valence state of each Ni0.5Mn0.5O2 layer. The obtained results are shown as vertical dashed lines in the right panel of Figure S6b, Supporting Information. Using the procedure described in the Experimental Section, we obtain an average valence state of the interfacial manganese atoms of +3.4, while a value of +3.1 is retrieved for the surface manganese atoms. At the interface, the fractional contributions of Mn3+ and Mn4+ are determined as 0.6 and 0.4, respectively. 2.4. Charge-Transfer Engineering via Top-Interface Design The electronic configuration obtained from both experimental and theoretical investigations of the LNMO//STO heterostructure is depicted in Figure4a. Electron-doped states triggered by the polar-discontinuity-induced oxygen vacancies are present at both interfaces. As these states are detrimental to the Mn4+–Ni2+-related LNMO ferromagnetism,[17,35] we want to control their extent. To this aim, we capped the LNMO layers with LNO. This compound is characterized by bulk paramagnetic metallic behavior over the whole temperature range[36] and its Ni3+ cation is a well-known electron acceptor.[2,3,5,37,38] Due to the strong electronegativity difference between Mn3+ and Ni3+,[17,38] in the LNO/LNMO//STO heterostructure, the effects of the polar discontinuity present at the LNMO bottom interface can potentially be relieved by an electronic reconstruction taking place at the LNMO top interface, toward the LNO side (Figure 4b). We verified the feasibility of this scenario by aligning the DFT density of states of LNMO and LNO via their work functions, as detailed in Figure S8, Supporting Information. The striking effect of LNO on the Mn/Ni electronic configuration of our LNMO//STO heterointerfaces is observed in the Mn and Ni L3,2-edge XAS spectra (Figure4c,d). The Mn3+ content present in a bare electron-doped 2 uc-thick LNMO is markedly reduced upon capping it with a 2 uc-thick LNO overlayer. Adv. Mater. 2022, 34, 2203071 Figure 3. a) HAADF image of a 13 uc LNMO//STO film. b) STEM-EDX map showing the distribution of Mn (green), Ni (red), and Ti (blue) in the LNMO//STO heterostructure, and schematic of the cation distribution in the octahedral sites. c) EELS of the O K and Mn L3,2 edges across the LNMO// STO heterostructure obtained at the positions indicated by the colored dots in (a). The intensity is normalized to the O Kb main peak and the Mn L3 line. Scale bars: 1nm. 15214095, 2022, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202203071 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [26/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License www.advmat.dewww.advancedsciencenews.com 2203071 (6 of 9) © 2022 The Authors. Advanced Materials published by Wiley-VCH GmbH The same effect is also obtained on a 5 uc-thick LNMO film (Figure S9a, Supporting Information). Such a drastic change in the Mn electronic configuration is not observed instead when the LNMO films are capped with LAO, as shown in Figure S9a, Supporting Information, and as predicted by our DFT calculations (Figure S8, Supporting Information). We can therefore exclude that in our LNO/LNMO//STO heterostructure, La vacancies (instead of Ni) are playing the role of electron acceptors.[40] Indeed, when the LNO overlayer is only 2 uc thick, the overall Ni L2-edge shape suggests that the Ni in LNO adopts a 2+ configuration (blue square and blue line in Figures 4b and 4d, respectively). A transition to more conventional Ni3+ (green square in Figure4b) is observed instead when the LNO thickness is increased (i.e., the case of 8 uc LNO capping shown by the green line in Figure 4d). The fact that the charge transfer is not observed when LAO is used as capping layer (the Al cation has no d electrons) highlights the pivotal role of Ni3+ in engineering such a functionality. We also verified the generality of this approach by characterizing LNO/LNMO//LSAT heterostructures with XAS (Figure S9b, Supporting Information). The observation of the same functional behavior on another polar interface further corroborates the ability of LNO to attract the electrons that are doping the opposite interface due to a polar discontinuity between the film and the substrate. As a result of this optimized electronic configuration, the magnetic properties of the LNO/LNMO//STO heterostructures are drastically enhanced. As shown in Figure 4e through the field-dependence of the Mn XMCD asymmetry measured at 20 K, the ferromagnetic behavior of the Mn sublattice in a 2-ucthick LNMO film is recovered upon LNO capping. The remarkable thickness-dependent evolution of the Curie temperature is summarized in Figure4f. We observe a systematic boost of TC in the top-interface-engineered heterostructures that are 5-ucthick or less. For example, an LNO-capped 2(5)-uc-thick LNMO is characterized by a TC≈ 80 (180) K as determined by combined XMCD and SQUID magnetometry (Figure S10, Supporting Information). This is to be compared with TC≈ 10 (165) K for a bare 2(5)-uc-thick LNMO film (Figure1d). For thicker LNMO Adv. Mater. 2022, 34, 2203071 Figure 4. a) Schematic of the charge configuration of the LNMO//STO heterostructure in presence of oxygen vacancies. Electrons transferred to the interfaces are depicted in yellow. b) New charge configuration after capping the LNMO//STO heterostructure with LNO. The doping electrons are now populating the first few LNO layers. In both (a) and (b), the colored squares highlight the location within the heterostructures of the indicated electronic configuration. c) XAS at Mn L3 edge of a 2 uc-thick LNMO film with and without top LNO. 10 uc-thick LNMO shown for comparison. Spectra shifted for better visibility. d) Same as (c) for the Ni L2 edge. The dashed line indicates the energy of the Ni L2-edge maximum of the reference LNMO sample. e) Mn XMCD asymmetry measured at 20 K for LNO/LNMO//STO (green circles) and LNMO//STO (red squares) heterostructures. Lines are a guide to the eye. f) Measured Curie temperature as a function of the LNMO thickness (log scale) for engineered (green circles) and bare (red squares) heterostructures. The data are fitted using a linear function and a power law for the capped heterostructures and a power law for the uncapped heterostructures.[39] The inset shows the low-thickness range using a linear scale. 15214095, 2022, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202203071 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [26/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License www.advmat.dewww.advancedsciencenews.com 2203071 (7 of 9) © 2022 The Authors. Advanced Materials published by Wiley-VCH GmbH films, the improvement is less evident given the reduced fraction that the interfacial layers contribute to the total film thickness and due to the increased distance between top and bottom interface that can inhibit the interlayer charge transfer.[41] Furthermore, it should be noted that in the bare LNMO// STO heterostructures, a conventional power law reduction of TC is observed in the whole thickness range. For the LNOengineered heterostructures, we can rather appreciate the presence of two different regimes for TC as a function of the LNMO thickness. The power-law reduction at larger thickness is followed by a linear behavior in the ultrathin domain (inset of Figure4f). This linear behavior is caused by the renormalization of the long-range spin-spin interaction due to finite-size effects.[39] By extrapolating the linear behavior to 0 uc LNMO, we obtain a TC value compatible with 0 K suggesting that the reduced dimensionality of the LNMO films is the only dominant factor determining the reduction of TC in the interfaceengineered LNO/LNMO//STO heterostructures. 3. Conclusion We have shown that ultrathin LNMO//STO heterostructures experience a degradation of their magnetic properties and poorer Ni/Mn cation ordering as a consequence of an oxygenvacancy-assisted electronic reconstruction taking place at LNMO polar interfaces. Introducing an LNO overlayer in these heterostructures redistributes the electron excess and restores the LNMO ferromagnetic properties. The LNO-induced electron migration is very general and can be adapted to other functional interfaces, being also obtained on heterostructures grown on LSAT, where a polarity mismatch takes place too. Such a toplayer engineering approach to manipulate the electronic state of a buried interface gets us closer to reaching a systematic and exact control of the desired functionalities in complex oxide heterostructures. 4. Experimental Section Sample Preparation and Characterization: LNO/LNMO//STO heterostructures were grown using off-axis radio-frequency magnetron sputtering in a flowing gas mixture of argon and oxygen stabilizing a pressure of 0.18 mbar. LNMO was grown at 720 °C while LNO was grown at 510 °C. The same conditions were used also on different substrates.[19] After the deposition, the heterostructures were cooled to room temperature in the same gas environment. Film thickness was calibrated via X-ray diffraction finite-size fringes[42] measured with a Rigaku Smartlab diffractometer operating in line focus mode with a monochromated Cu Kα1 radiation. Magnetic measurements were carried out in a Quantum Design superconducting quantum interference device (SQUID) magnetometer operating in vibrating sample mode (VSM) with the magnetic field applied parallel to the plane of the sample. A residual magnetic moment from the STO substrates has been subtracted. Curie temperatures were determined from SQUID magnetometry taking the minimum of the derivative of the magnetic moment. X-ray Absorption Spectroscopy (XAS) and X-ray Magnetic Circular Dichroism (XMCD): Measurements were performed at both the X-Treme beamline[43] at the Swiss Light Source (SLS) and BOREAS beamline[44] at the ALBA Synchrotron. The spectra were acquired by measuring the total electron yield with an incidence angle (the angle between the X-ray beam direction and the sample normal) of 60°. The XAS curves were obtained as the sum of spectra measured with vertical and horizontal linear polarization and were normalized to the peak of the respective L3,2-edge. The reference spectra were obtained from MnCl2 (Mn2+), Mn2O3 (Mn3+), SrMnO3 (Mn4+), Nd0.8Sr0.2NiO2+δ (Ni1.5+), NiO (Ni2+), and NdNiO3 (Ni3+) and were normalized to the peak of the respective L3-edge. The Nd0.8Sr0.2NiO2+δ sample was obtained using a topotactic reduction.[45] Only for this specific case, the effective Ni valence was inferred by comparison with charge transfer multiplet theory calculations.[46] The XMCD was defined as the difference between spectra measured with left and right circular polarization and normalized to the L3 maximum from the XAS. The XMCD asymmetry was defined as / ()() −+ LR LR with L(R) obtained as the difference between the L3-resonant and off-resonant energies for left (right) circular polarization.[19] The XMCD Ni L3 integral values as a function of the temperature presented in Figure1d have been obtained in an external magnetic field of 0.5 T for the 5 and 3 uc-thick LNMO films, and in an external magnetic field of 5 T for the 2 uc-thick film. This difference in magnetic field intensity causes an increase in TC of around 10 K.[19] This information has been used for a more realistic determination of TC in the thinner LNMO sample. Density Functional Theory (DFT): DFT calculations were performed using the PBEsol exchange-correlation functional as implemented in the VASP package.[47] Electron–core interactions were described via the projector augmented wave (PAW) method[48] with La(5s, 5p, 5d, 6s), Mn(3p, 3d, 4s), Ni(3p, 3d, 4s), Sr(4s, 4p, 5s), Ti(3s, 3p, 3d, 4s), and Ga(3d, 4s, 4p). Valence shells and wave functions were expanded in a plane waves basis with 500eV kinetic energy cutoff. A Hubbard U correction[49] was applied to the d states of Mn (3.0eV), Ni (6.0 eV), and Ti (4.0eV). The interface was modeled as cell with 22× in-plane dimension (5.454 × 5.454 Å) with respect to the 5-atom cubic cell. In the out-of-plane direction, 4 uc layers of STO and 6 uc layers of LNMO were used together with a vacuum of 12 Å. Reciprocal space was sampled using 6 × 6 × 1 mesh. A ferromagnetic spin configuration was initialized for Mn and Ni atoms in the thin film. Structures were relaxed until forces converged below 0.01eV Å−1. Scanning Transmission Electron Microscopy (STEM): Electron transparent cross-sectional samples for transmission electron microscopy were prepared by means of an FEI Helios NanoLab 600i focused ion beam operated at accelerating voltages of 30 and 5 kV. High-angle annular dark-field STEM (HAADF-STEM), EDX spectroscopy, and EELS were carried out using a probe-corrected FEI Titan Themis microscope equipped with ChemiSTEM technology and a CEOS Energy-Filtering and Imaging Device (CEFID) in combination with a direct electron detector (ELA, Dectris). The microscope was operated at an accelerating voltage of 300 kV. For the HAADF and EDX data acquisition, a probe convergence semiangle of 18 mrad was set and the annular semidetection range of the annular dark-field detector was calibrated at 84–200 mrad. The elemental maps were calculated from the EDX spectrum image using the Sr Kα, Ti Kα, La Lα, Mn Kα, and Ni Kα lines. Note that even though the EDX spectrum image in Figure3b was performed with a low beam current of 80 pA, the two topmost Ni0.5Mn0.5O2 layers were sputtered away during the 6.6 min long acquisition. The EELS data were obtained by setting the convergence and collection semiangles to 26 and 35 mrad, respectively, yielding an effective collection semiangle of about 29.5 mrad for the used energyloss range. An energy dispersion of 0.19eV per channel was chosen to simultaneously record both the O K and Mn L3,2 edges. A 390 × 150pixel EEL spectrum image (see Figure S6, Supporting Information) was acquired with a pixel dwell time of 0.5ms, resulting in a total acquisition time of 30s. Note that the electron probe was focused on the film for only about half of the time, while the rest of the time it was focused on the substrate. The low beam current used (80 pA) in combination with the extremely short acquisition time prevented film damage from electron beam irradiation. This was possible thanks to the large CEFID entrance aperture used along with the high sensitivity of the ELA detector. All spectra were background subtracted by fitting a decaying power-law function to an energy window just in front of the core-loss edge onsets. The EEL spectra shown in Figure 3c were calculated by Adv. Mater. 2022, 34, 2203071 15214095, 2022, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202203071 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [26/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License www.advmat.dewww.advancedsciencenews.com 2203071 (8 of 9) © 2022 The Authors. Advanced Materials published by Wiley-VCH GmbH Adv. Mater. 2022, 34, 2203071 summing all the spectra at each individual Ni0.5Mn0.5O2 slab. The local Mn valence state of each Ni0.5Mn0.5O2 layer was quantified from the Mn L3,2 white-line intensity ratio after background subtraction using scaled Hartree-Slater cross-sections.[33,50,51] For the determination of the Mn oxidation state, reference spectra for Mn4+ and Mn3+ were acquired from CaMnO3 and YMnO3, respectively. The valence state of the manganese atoms was estimated by fitting a linear regression to the experimental Mn L3/L2 intensity ratio of the two reference spectra over the manganese valence state. The linear regression produced the following equation: y= 3.85 − 0.47x (where x is the manganese valence state and y is the Mn L3/L2 ratio). Supporting Information Supporting Information is available from the Wiley Online Library or from the author. Acknowledgements This research was supported by the Swiss National Science Foundation (SNSF) under Project No. PP00P2_170564 and R’Equip Project No. 206021_150784 ASKUZI. S.J. and M.G. acknowledge the financial support of Agility plus grant from MARVEL NCCR (SNSF Project No. 51NF40182892). M.K. and U.A. were supported by the SNSF, Project No. 200021_178791. M.C, R.E., and M.D.R. acknowledge the financial support of the SNSF, R’Equip Project No. 206021_189625. A.Z. acknowledges the financial support of the SNSF, Project No. 200021_169467. G.D.L., J.S., S.J., and M.G. thank Thomas Greber for granting SQUID access. DFT calculations were performed on UBELIX (http://www.id.unibe.ch/hpc), the HPC cluster at the University of Bern and on Piz Daint at the Swiss Supercomputing Center CSCS under projects s955 and s1033. Open access funding provided by Universitat Zurich. Conflict of Interest The authors declare no conflict of interest. Author Contributions G.D.L. and M.G. conceived the project and designed the experiments. G.D.L., J.S., and S.J. grew and characterized the heterostructures and performed SQUID magnetometry, XAS, and XMCD measurements. Synchrotron measurements were supported by A.Z. and C.P. at SLS and C.G. and J.H. at ALBA and were analyzed by G.D.L. TEM specimens were prepared and characterized by M.C., R.E., and M.D.R. DFT calculations were performed by M.K. under the supervision of U.A. First draft of the manuscript was written by G.D.L. and M.G. All authors analyzed the data, discussed the results, and contributed to the final version of the paper. Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request. Keywords charge-transfer phenomena, double perovskites, ferromagnetism, oxide interfaces Received: April 5, 2022 Revised: July 12, 2022 Published online: August 7, 2022 [1] A.Ohtomo, H. Y.Hwang, Nature 2004, 427, 423. [2] A. J.Grutter, H.Yang, B. J.Kirby, M. R.Fitzsimmons, J. A.Aguiar, N. D.Browning, C. A.Jenkins, E.Arenholz, V. V.Mehta, U. S.Alaan, Y.Suzuki, Phys. Rev. Lett. 2013, 111, 087202. [3] M.Gibert, P.Zubko, R.Scherwitzl, J.Íñiguez, J.-M.Triscone, Nat. Mater. 2012, 11, 195. [4] J. Chakhalian, J. W. Freeland, H.-U. Habermeier, G. Cristiani, G.Khaliullin, M.vanVeenendaal, B.Keimer, Science 2007, 318, 1114. [5] A. S. Disa, D. P. Kumah, A. Malashevich, H. Chen, D. A. Arena, E. D.Specht, S.Ismail-Beigi, F. J.Walker, C. H.Ahn, Phys. Rev. Lett. 2015, 114, 026801. [6] Z.Zhong, P.Hansmann, Phys. Rev. X 2017, 7, 011023. [7] H. Yamada, Y. Ogawa, Y. Ishii, H. Sato, M. Kawasaki, H. Akoh, Y.Tokura, Science 2004, 305, 646. [8] M.Nord, P. E.Vullum, M.Moreau, J. E.Boschker, S. M.Selbach, R.Holmestad, T.Tybell, Appl. Phys. Lett. 2015, 106, 041604. [9] H. Boschker, J. Verbeeck, R. Egoavil, S. Bals, G. van Tendeloo, M. Huijben, E. P. Houwman, G. Koster, D. H. A. a Blank, G.Rijnders, Adv. Funct. Mater. 2012, 22, 2235. [10] D.Kan, R.Aso, R.Sato, M.Haruta, H.Kurata, Y.Shimakawa, Nat. Mater. 2016, 15, 432. [11] Z.Liao, M.Huijben, Z.Zhong, N.Gauquelin, S.Macke, R. J.Green, S.Van Aert, J.Verbeeck, G.Van Tendeloo, K.Held, G. A.Sawatzky, G.Koster, G.Rijnders, Nat. Mater. 2016, 15, 425. [12] A. D. Caviglia, S. Gariglio, N. Reyren, D. Jaccard, T. Schneider, M. Gabay, S. Thiel, G. Hammerl, J. Mannhart, J.-M. Triscone, Nature 2008, 456, 624. [13] P.Noël, F.Trier, L. M.Vicente Arche, J.Bréhin, D. C.Vaz, V.Garcia, S.Fusil, A.Barthélémy, L.Vila, M.Bibes, J. P.Attané, Nature 2020, 580, 483. [14] S.Wang, Y.Bai, L.Xie, C.Li, J. D.Key, D.Wu, P.Wang, X.Pan, ACS Appl. Mater. Interfaces 2018, 10, 1374. [15] M. J.Chen, X. K.Ning, Z. J.Wang, P.Liu, S. F.Wang, J. L.Wang, G. S.Fu, S. Ma, W.Liu, Z. D. Zhang, Appl. Phys. Lett. 2018, 112, 021601. [16] N.Nakagawa, H. Y.Hwang, D. A.Muller, Nat. Mater. 2006, 5, 204. [17] R. I. Dass, J.-Q. Yan, J. B. Goodenough, Phys. Rev. B 2003, 68, 064415. [18] N. S.Rogado, J.Li, A. W.Sleight, M. A.Subramanian, Adv. Mater. 2005, 17, 2225. [19] G. De Luca, J. Spring, U. Bashir, M. Campanini, R. Totani, C.Dominguez, A.Zakharova, M.Döbeli, T.Greber, M. D.Rossell, C.Piamonteze, M.Gibert, APL Mater. 2021, 9, 081111. [20] M. C.Sánchez, J.García, J.Blasco, G.Subías, J.Perez-Cacho, Phys. Rev. B 2002, 65, 144409. [21] U.Aschauer, R.Pfenninger, S. M.Selbach, T.Grande, N. A.Spaldin, Phys. Rev. B 2013, 88, 054111. [22] B. Gilbert, B. H. Frazer, A. Belz, P. G. Conrad, K. H. Nealson, D.Haskel, J. C.Lang, G.Srajer, G.De Stasio, J. Phys. Chem. A 2003, 107, 2839. [23] H.Wang, D. S.Patil, W.Gu, L.Jacquamet, S.Friedrich, T.Funk, S. P. Cramer, J. Electron Spectrosc. Relat. Phenom. 2001, 114–116, 855. [24] X. R.Wang, C. J.Li, W. M.Lu, T. R.Paudel, D. P.Leusink, M.Hoek, N.Poccia, A.Vailionis, T.Venkatesan, J. M. D.Coey, E. Y.Tsymbal, Ariando, H.Hilgenkamp, Science 2015, 349, 716. [25] Z. Chen, Z. Chen, Z. Q. Liu, M. E. Holtz, C. J. Li, X. R. Wang, W. M.Lü, M.Motapothula, L. S.Fan, J. A.Turcaud, L. R.Dedon, C.Frederick, R. J.Xu, R.Gao, A. T.N’Diaye, E.Arenholz, J. A.Mundy, 15214095, 2022, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202203071 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [26/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License www.advmat.dewww.advancedsciencenews.com 2203071 (9 of 9) © 2022 The Authors. Advanced Materials published by Wiley-VCH GmbH Adv. Mater. 2022, 34, 2203071 T.Venkatesan, D. A.Muller, L.-W.Wang, J.Liu, L. W.Martin, Phys. Rev. Lett. 2017, 119, 156801. [26] L.Yu, A.Zunger, Nat. Commun. 2014, 5, 5118. [27] N. C.Bristowe, P. B.Littlewood, E.Artacho, Phys. Rev. B 2011, 83, 205405. [28] Z. Huang, K. Han, S. Zeng, M. Motapothula, A. Y. Borisevich, S.Ghosh, W.Lü, C. Li, W.Zhou, Z. Liu, M. Coey, T. Venkatesan, Ariando, Nano Lett. 2016, 16, 2307. [29] Z. Wang, A. H. Tavabi, L. Jin, J. Rusz, D. Tyutyunnikov, H. Jiang, Y. Moritomo, J. Mayer, R. E. Dunin-Borkowski, R. Yu, J. Zhu, X.Zhong, Nat. Mater. 2018, 17, 221. [30] S. A.Chambers, L. Qiao, T. C.Droubay, T. C.Kaspar, B. W. Arey, P. V.Sushko, Phys. Rev. Lett. 2011, 107, 4. [31] J. A.Mundy, Y.Hikita, T.Hidaka, T.Yajima, T.Higuchi, H. Y.Hwang, D. A.Muller, L. F.Kourkoutis, Nat. Commun. 2014, 5, 3464. [32] T. C. Kaspar, P. V. Sushko, S. R. Spurgeon, M. E. Bowden, D. J.Keavney, R. B.Comes, S.Saremi, L.Martin, S. A.Chambers, Adv. Mater. Interfaces 2019, 6, 1801428. [33] M.Varela, M. P.Oxley, W.Luo, J.Tao, M.Watanabe, A. R.Lupini, S. T.Pantelides, S. J.Pennycook, Phys. Rev. B 2009, 79, 085117. [34] P. Agrawal, J. Guo, P. Yu, C. Hébert, D. Passerone, R. Erni, M. D.Rossell, Phys. Rev. B 2016, 94, 104101. [35] M. P.Singh, K. D.Truong, S.Jandl, P.Fournier, J. Appl. Phys. 2010, 107, 09D917. [36] S.Catalano, M.Gibert, J.Fowlie, J.Íñiguez, J.-M.Triscone, J.Kreisel, Rep. Prog. Phys. 2018, 81, 046501. [37] H. Chen, A. J. Millis, C. A. Marianetti, Phys. Rev. Lett. 2013, 111, 116403. [38] C. Piamonteze, M. Gibert, J. Heidler, J. Dreiser, S. Rusponi, H.Brune, J.-M. M.Triscone, F.Nolting, U.Staub, Phys. Rev. B 2015, 92, 014426. [39] R.Zhang, R. F.Willis, Phys. Rev. Lett. 2001, 86, 2665. [40] L.Wu, C. Li, M.Chen, Y.Zhang, K.Han, S. Zeng, X. Liu, J.Ma, C.Liu, J.Chen, J.Zhang, Ariando, T. V.Venkatesan, S. J.Pennycook, J. M. D.Coey, L.Shen, J.Ma, X. R. Wang, C. W. Nan, ACS Appl. Mater. Interfaces 2017, 9, 44931. [41] G. J. Omar, M. Li, X. Chi, Z. Huang, Z. S. Lim, S. Prakash, S.Zeng, C.Li, X.Yu, C.Tang, D.Song, A.Rusydi, T.Venkatesan, S. J.Pennycook, A.Ariando, Nano Lett. 2020, 20, 2493. [42] C.Lichtensteiger, J. Appl. Crystallogr. 2018, 51, 1745. [43] C. Piamonteze, U. Flechsig, S. Rusponi, J. Dreiser, J. Heidler, M. Schmidt, R. Wetter, M. Calvi, T. Schmidt, H. Pruchova, J. Krempasky, C. Quitmann, H. Brune, F. Nolting, J. Synchrotron Radiat. 2012, 19, 661. [44] A.Barla, J.Nicolás, D.Cocco, S. M.Valvidares, J.Herrero-Martín, P.Gargiani, J.Moldes, C.Ruget, E.Pellegrin, S.Ferrer, J. Synchrotron Radiat. 2016, 23, 1507. [45] D.Li, K.Lee, B. Y.Wang, M.Osada, S.Crossley, H. R.Lee, Y.Cui, Y.Hikita, H. Y.Hwang, Nature 2019, 572, 624. [46] F. M. F.deGroot, J. Electron Spectrosc. Relat. Phenom. 1994, 67, 529. [47] G.Kresse, J. Non-Cryst. Solids 1995, 192–193, 222. [48] P. E.Blöchl, Phys. Rev. B 1994, 50, 17953. [49] S. L. Dudarev, G. A. Botton, S. Y. Savrasov, C. J. Humphreys, A. P.Sutton, Phys. Rev. B 1998, 57, 1505. [50] T.Riedl, T.Gemming, K.Wetzig, Ultramicroscopy 2006, 106, 284. [51] H. K.Schmid, W.Mader, Micron 2006, 37, 426. 15214095, 2022, 36, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/adma.202203071 by Csic Organización Central Om (Oficialia Mayor) (Urici), Wiley Online Library on [26/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License