Research Article www.acsami.org
Cite This: ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Cross-Correlation between Strain, Ferroelectricity, and Ferromagnetism in Epitaxial Multiferroic CoFe2O4/BaTiO3 Heterostructures Nathalie Jedrecy,*,† Thomas Aghavnian,‡ Jean-Baptiste Moussy,‡ Hélène Magnan,‡ Dana Stanescu,‡ Xavier Portier,§ Marie-Anne Arrio,∥ Cristian Mocuta,⊥ Alina Vlad,⊥ Rachid Belkhou,⊥ Philippe Ohresser,⊥ and Antoine Barbier‡ Downloaded via DURHAM UNIV on August 8, 2018 at 06:12:28 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
†
Institut des Nano Sciences de Paris (INSP), Sorbonne Université, CNRS UMR 7588, 4 Place Jussieu, 75252 Paris Cedex 05, France Service de Physique de l’Etat Condensé (SPEC), CEA, CNRS UMR 3680, Université Paris Saclay, Orme des Merisiers, CEA Saclay, 91191 Gif sur Yvette Cedex, France § Centre de recherche sur les Ions, les MAtériaux et la Photonique (CIMAP), CEA, CNRS UMR 6252, ENSICAEN, Normandie Université, 6 Boulevard Maréchal Juin, 14050 Caen, France ∥ Institut de Minéralogie, de Physique des Matériaux et de Cosmochimie (IMPMC), Sorbonne Université, CNRS UMR 7590, IRD, MNHN, 4 Place Jussieu, 75252 Paris Cedex 05, France ⊥ Synchrotron SOLEIL, L’Orme des Merisiers Saint-Aubin, BP 48, 91192 Gif sur Yvette Cedex, France ‡
ABSTRACT: Multiferroic biphase systems with robust ferromagnetic and ferroelectric response at room temperature would be ideally suitable for voltage-controlled nonvolatile memories. Understanding the role of strain and charges at interfaces is central for an accurate control of the ferroelectricity as well as of the ferromagnetism. In this paper, we probe the relationship between the strain and the ferromagnetic/ferroelectric properties in the layered CoFe2O4/BaTiO3 (CFO/BTO) model system. For this purpose, ultrathin epitaxial bilayers, ranging from highly strained to fully relaxed, were grown by molecular beam epitaxy on Nb:SrTiO3(001). The lattice characteristics, determined by X-ray diffraction, evidence a non-intuitive crosscorrelation: the strain in the bottom BTO layer depends on the thickness of the top CFO layer and vice versa. Plastic deformation participates in the relaxation process through dislocations at both interfaces, revealed by electron microscopy. Importantly, the switching of the BTO ferroelectric polarization, probed by piezoresponse force microscopy, is found dependent on the CFO thickness: the larger is the latter, the easiest is the BTO switching. In the thinnest thickness regime, the tetragonality of BTO and CFO has a strong impact on the 3d electronic levels of the different cations, which were probed by Xray linear dichroism. The quantitative determination of the nature and repartition of the magnetic ions in CFO, as well as of their magnetic moments, has been carried out by X-ray magnetic circular dichroism, with the support of multiplet calculations. While bulklike ferrimagnetism is found for 5−15 nm thick CFO layers with a magnetization resulting as expected from the Co2+ ions alone, important changes occur at the interface with BTO over a thickness of 2−3 nm because of the formation of Fe2+ and Co3+ ions. This oxidoreduction process at the interface has strong implications concerning the mechanisms of polarity compensation and coupling in multiferroic heterostructures. KEYWORDS: multiferroics, strain, ferroelectric polarization, X-ray absorption, X-ray magnetic circular dichroism, ferrimagnetism, oxide interface
1. INTRODUCTION
as for the development of next generation nonvolatile magnetic memories addressable by an electric field.5−10 Indeed, if both ferroic orders are cross-coupled, one can design magnetoelectric (ME) random access memories, combining energy efficiency and fast switching speed, with electric write and magnetic read of
In crystalline transition metal oxides, the lattice, charge, orbital, and spin degrees of freedom and their interplay allow us to consider a rich variety of functional properties that may be combined for new performances in applications.1−4 In that respect, multiferroic oxides in which charge and spin long-range orders coexist, leading respectively to ferroelectric (FE) and ferromagnetic (FM) (or antiferromagnetic) properties, are extensively studied from a fundamental point of view, as well © XXXX American Chemical Society
Received: June 12, 2018 Accepted: July 23, 2018
A
DOI: 10.1021/acsami.8b09499 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Research Article
ACS Applied Materials & Interfaces
investigated. Unexpectedly, we find that the strain state and the electric polarization in BTO strongly depend on the thickness of the CFO layer on top. In addition, we give evidence of a chemically modified CFO region close to the BTO interface, with the formation of Fe2+ and Co3+ ions. As a consequence, while the average magnetic moment of Fe increases with decreasing CFO thickness, that of Co decreases. Bulklike FM features are restored above 4.4 nm of thickness. Our results show that very different FE, FM, or tunneling behaviors may be obtained from a bilayer system, depending on the relative thicknesses of the two types of layers and on their interface.
the logic states.11 The lack of single-phase oxides with robust multiferroicity at room temperature has oriented most of the investigations toward artificial multiferroic systems based on the simultaneous presence of two phases, one FE and the other one FM. These two-phase multiferroic materials may take the form of either vertical self-assembled nanocomposites or horizontal heterostructures.12−16 In each case, the interfaces between the two phases will play a major role in the final properties and in the cross-coupling between the ferroic orders. A prototype system is BaTiO3 (BTO)/CoFe2O4 (CFO) where BTO is a typical FE of the displacive type below 395 K and CFO a ferromagnet with a high Curie temperature (793 K).12,16−19 Strictly speaking, CFO belongs to the class of ferromagnets named ferrimagnets, where antiferromagetic interactions by super-exchange between magnetic moments of different magnitudes lead to a finite magnetization. For the sake of simplicity, we will refer to CFO as a ferromagnet. In vertical nanocomposite films, where FM CFO nanopillars are embedded in a FE BTO matrix, strain-mediated ME interactions have been claimed.12 The large number of parameters acting on the ME behavior, such as the degree of regularity of the pillar network, the distance between the nanopillars, the exact chemical nature of the various CFO/BTO interfaces, do not allow a detailed understanding of the strain-mediated coupling mechanisms. Horizontal heterostructures are more suitable systems for the investigation of the individual physical properties, in addition to being more relevant for device fabrication.18−20 The modulation by lattice strain of the dielectric, FE, and FM properties in epitaxial bilayers has been evidenced for PZT/CFO.20 In BTO/ CFO and CFO/BTO layered systems, ME effects have been reported, consisting in changes of the permittivity of BTO by a magnetic field18 and in the measurement of ME voltage coefficients.19 However, no systematic investigation of both FE and FM properties (including determination of atomic magnetic moments) as a function of the thickness of each layer (CFO or BTO), nor of the nature of the CFO/BTO interface at the nanometer scale, has been carried out. Besides the scope of multiferroics, layered systems involving either a FM or a FE component are involved in a large variety of spintronic devices, from spin injectors to magnetic or FE tunnel junctions.3 Large magnetoresistive effects have been obtained with CFO-based epitaxial multilayers.21 The BTO-based tunnel junctions are on their side extensively studied for their giant tunneling electroresistance, when switching the polarization of the FE barrier.22,23 It is recognized that progress in the field of tunnel junctions relies on a deeper understanding of the structure and electronic states of each type of layer and of the interfaces in the ultrathin thickness regime. In this work, we study a panel of horizontal CFO/BTO heterostructures grown by atomic oxygen-assisted molecular beam epitaxy (AO-MBE) on Nb(1%)-doped SrTiO3 (STO) (001) substrates. The CFO and BTO layers, with individual thicknesses from 2 to 15 nm, range from highly strained with respect to the substrate to fully relaxed. We analyze in detail the interrelation between the strain and the FE/FM properties of the full stack, highlighting the roles of the two interfaces: BTO/ substrate and CFO/BTO. A combination of macroscopic and microscopic measurements, in particular, by advanced X-ray synchrotron-based spectroscopic techniques with atomic sensitivity, is used to follow the evolution of the structural and spin-resolved electronic properties of the bottom BTO and top CFO layers. In particular, the enhanced tetragonal state of the two types of layers in the thinnest films is thoroughly
2. EXPERIMENTAL DETAILS 2.1. Film Growth. The CFO/BTO heterostructures were grown by AO-MBE on Nb:STO(001) substrates pre-annealed in air. Nb-doped conductive substrates were used for serving as ground electrodes in the piezoresponse measurements and for avoiding any charging effects during the X-ray absorption spectroscopy experiments. The growth setup works in the 10−10 mbar range and is equipped with a radiofrequency oxygen plasma source. Prior to deposition, the substrates were first outgassed at 900 °C and then exposed to a high brilliance oxygen plasma (power 350 W) at 450 °C. The metal species were then evaporated under the same conditions from dedicated Knudsen cells (purity 99.99%). The BTO layers were grown first, followed by CFO. The growth rates, predetermined by X-ray reflectivity, are around 0.117 nm min−1 for BTO and 0.044 nm min−1 for CFO. The oxide layers are fully oxidized and stoichiometric; this was checked by Auger electron spectroscopy and X-ray photoelectron spectroscopy. The structural quality of the films, in particular, the bidimensional growth mode at the early stages, was controlled in situ during the growth by reflection highenergy electron diffraction. 2.2. Structural Analysis. The X-ray diffraction (XRD) measurements were carried out at the SixS and DiffAbs beamlines of the SOLEIL synchrotron (Gif-sur-Yvette, France). The two beamlines supply monochromatic photons in the 5−20 and 3−25 keV energy (E) range with a ΔE/E resolution of 10−4 and are equipped with a 2 + 3- and 6-circle diffractometer, respectively. The reciprocal space mapping (RSM) was conducted using a photon energy of 15 keV in grazing incidence geometry. The microstructure of some films was analyzed by scanning transmission electron microscopy (STEM) using a high angle angular dark field (HAADF) detector. The DigitalMicrograph software of GATAN was used for the fast Fourier transformation (FFT). Chemical maps were recorded by energy dispersive X-ray (EDX) spectroscopy combined with the STEM mode of the electron microscope. For these observations, a double-corrected cold FEG ARM200F JEOL microscope operated at 200 kV and equipped with a Centurio EDX JEOL setup was used. TEM foils were prepared by a focused ion beam system (FEI HELIOS Nanolab 660). 2.3. Spectroscopic Analysis. The X-ray absorption spectroscopy (XAS) and dichroic measurements were performed at the DEIMOS beamline (SOLEIL synchrotron). This beamline is dedicated to soft Xray dichroic measurements in the 350−2500 eV photon energy range.24 Two undulators are used to provide fully circular or fully linear polarized light with a fast switching between the left and right or between the vertical and horizontal configurations. The end-station is equipped with a cryo-magnet (magnetic field up to 7 T along the beam); the sample temperature may be varied from 1.5 to 370 K. The XAS spectra were measured using total electron yield detection. 2.4. Piezoresponse Measurements. The atomic force microscopy and piezoresponse force microscopy (PFM) images were obtained by use of a Brüker Dimension ICON microscope under Nanoscope V controller from CEA-SPEC.
3. RESULTS AND DISCUSSION 3.1. Structural Analysis. At room temperature, the perovskite BTO oxide has a tetragonal structure with bulk lattice cell parameters aBTO = 3.992 Å and cBTO = 4.036 Å, while B
DOI: 10.1021/acsami.8b09499 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Research Article
ACS Applied Materials & Interfaces
Figure 1. (a) RSM close to the 222STO reflection for the 2.2 nm/5 nm CFO/BTO film and (b) for the 15 nm/10 nm CFO/BTO film. (c) RSM close to 222STO for the 10 nm/10 nm CFO/BTO film. Two regions are considered corresponding to a relaxed and to a strained part of the BTO layer. (d) Inplane (aBTO) and out-of-plane (cBTO) lattice parameters in BTO as a function of the CFO thickness for a 10 nm thick BTO film. Strained (filled symbols) and relaxed (opened symbols) regions are considered, close to the BTO/STO and CFO/BTO interfaces, respectively. The thick continuous line is the mean value, based on the respective proportions of each region. The horizontal dotted lines figure the bulk BTO values. (e) Corresponding lattice parameters in CFO. The in-plane lattice unit cells of BTO and CFO are sketched for the 4 main strain states (green CFO, red BTO, and black STO).
the STO perovskite oxide is cubic with aSTO = 3.905 Å. The lattice mismatch, defined as (afilm − asubs)/asubs, is +2.2% and in the low film thickness range, the substrate will generate a compressive biaxial in-plane stress of the BTO film. The CFO oxide has a spinel cubic structure with aCFO = 8.392 Å. Considering half the unit cell, the lattice mismatch with respect to the STO substrate is +7.4%, while that with respect to BTO is +5.1%. The CFO film is expected to be in compressive in-plane stress at the early stages of the growth, while the BTO film will be submitted to tensile stress by CFO. The epitaxial growth of the CFO/BTO heterostructure was performed by AO-MBE starting from STO(001) substrates. Details on the growth conditions can be found in refs.25−27 Several samples were produced in the 5−15 nm thickness range for BTO and in the 2.2−15 nm range for CFO. The BTO and CFO layers grow with the (001) orientation and a cube-on-cube epitaxy, with parallel in-plane [100] axes. The two types of oxide layers are fully oxidized (without detectable oxygen vacancies) and stoichiometric; the inversion parameter of the CFO spinel is generally about 0.85.27 Two typical RSMs obtained by XRD about the 222 reflection of the substrate are shown in Figure 1a,b. They correspond to the thinnest and thickest stack investigated, respectively. The HKL indexes are expressed in reciprocal lattice units of the STO lattice. As seen in Figure 1b, for the thickest film (15 nm of CFO on 10 nm BTO), the 222BTO and 444CFO reflections appear distinctly at lower H and L indexes than 222STO. The first peak (H ≅ L = 1.957) is related to BTO while the second peak (H ≅ L = 1.871) is related to CFO. The two types of layers are almost fully relaxed with average lattice parameters close to their bulk values: a ≅ c = 3.99 Å for BTO and a ≅ c = 8.35 Å for CFO. The
BTO layer appears here in a nearly cubic form because of the thick CFO overlayer with its larger in-plane lattice spacing, which counterbalances the complete relaxation of BTO toward its bulk tetragonal form. The intensities of 222BTO and 444CFO are both distributed according to inclined ellipses in the HL plane, attesting the existence of a strain gradient in each layer. For the thinnest film (2.2 nm of CFO on 5 nm BTO), the intensity is distributed along the out-of-plane direction (the ellipse is quasi vertical) at fixed H ≅ 2 value with a maximum at about L = 1.9, as seen in Figure 1a. The BTO film is quasicoherent with respect to the substrate, the aBTO and cBTO mean lattice parameters being about 3.915 and 4.11 Å, respectively. If we take into account the dispersion of the X-ray diffracted intensity in the out-of-plane and in-plane directions, values close to 3.905 Å for aBTO and 4.34 Å for cBTO may be inferred at the BTO/STO interface, the c/a ratio reaching therein 1.11 (the bulk c/a ratio is 1.011). The CFO overlayer is hardly distinguishable in the 222 map but a strained tetragonal state has been deduced from the 202 RSM measurement. At intermediate thicknesses, the BTO film has to be described as divided in two differently strained regions: a region near the STO interface in a highly in-plane compressed state and a region near the CFO interface in a relaxed state or even in potential tensile stress. The existence of these two regions is revealed by the intensity distribution in the RSMs. Figure 1c shows the situation for 10 nm BTO and 10 nm CFO on top. The intensity of 222BTO is distributed according to an ellipse whose center does not coincide with the intensity maximum, which is displaced toward the H = 2 vertical line. The center of the ellipse was chosen for evaluating the a and c average lattice parameters C
DOI: 10.1021/acsami.8b09499 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Research Article
ACS Applied Materials & Interfaces
deposits. Considering the growing interest in ultrathin BTO films with a few unit cells in FE tunnel junctions22,31 or in multiferroic heterostructures, the precise knowledge of all factors which may affect the energy barrier profile and the future FE response, among which is the strain state, is of crucial importance. Lastly, the CFO layer appears much more elastic than BTO and the strain generated in BTO by the clamping to the STO substrate is efficiently transmitted to the CFO overlayer. This is reflected in the larger c/a ratio reached for CFO in the biaxial compressive state with respect to BTO. For the same elastic reasons, we observe a faster decrease of the c and a values in CFO compared to BTO; the heterostructure strain is released primarily in the CFO overlayer. Epitaxial CFO films may easily be grown with either tensile (c/a < 1) or compressive (c/a > 1) strain, and a related complete switching of the magnetic anisotropy axis has been reported.32 Our measurements in the ultrathin range (2−6 nm) show that a very marked tetragonality (c/a = 1.08) may indeed be reached in CFO. The strain plays a critical role in tuning both the dielectric33 and the magnetic34 properties of CFO. The relaxation scheme in the CFO/BTO heterostructure is confirmed by the high-resolution transmission electron microscopy analysis. Figure 2a shows a cross-sectional image
in the relaxed region, while the position of the intensity maximum was retained as indicative of the lattice in the strained region. The RSMs, performed around 222STO and 402STO for each sample and analyzed this way, allowed us to extract the evolution of the a (in-plane) and c (out-of-plane) lattice parameters in BTO as a function of the CFO thickness. Three fixed BTO thicknesses (5, 10, 15 nm) were investigated and the result for the 10 nm BTO bottom layer is shown in Figure 1d. The BTO layer evolves from highly tetragonal (c/a = 1.05) to nearly cubic (c/a = 1.005) when the thickness of CFO increases. The region near the STO interface stays tetragonal up to ≅10 nm of CFO on top, while a relaxation close to the CFO interface occurs as soon as 2.2 nm of CFO is deposited. The volume of the relaxed BTO regions with respect to the strained regions increases with increasing CFO thickness on top, and the mean c and a values are expected to evolve according to the two intermediate thick lines sketched in Figure 1d. The corresponding CFO lattice parameters are shown in Figure 1e. The CFO overlayer is highly tetragonal (c/a = 1.074) in the low thickness range as a result of the strained state of BTO beneath and it relaxes progressively toward its natural cubic form for higher thicknesses. One may distinguish four strain states of the CFO/ BTO films. First, at the very beginning of the CFO growth, the BTO and CFO layers are compressively stressed by the STO substrate. Second, with increasing CFO thickness, the CFO layer progressively relaxes while the BTO keeps strained with respect to STO. Third, the BTO bottom layer relaxes. Fourth, the two types of layers nearly reach their bulk state. Considering a fixed BTO thickness of 5 nm, we found the same tendencies, except the fact that the tetragonality of BTO persists up to 15 nm of CFO, the mean c/a ratio in BTO evolving from 1.046 to 1.015 (from 2.2 to 15 nm of CFO). In parallel, the CFO layer remains tetragonal longer and at 10 nm, its c/a ratio is 1.021. The (intensity weighted) mean values of the c/a ratio for the two types of layer as a function of the CFO thickness considering the 10 and 5 nm thick BTO layers are given in Table 1.
Figure 2. (a) Cross-sectional STEM-HAADF images of a 4.4 nm/10 nm CFO/BTO film on STO(001), and the corresponding STEM− EDX chemical maps of the Sr, Ti, Ba, O, Co, and Fe elements. The interface is along [100]STO. (b) High-resolution STEM-HAADF and filtered images of the same film giving evidence of dislocations, in the BTO layer (red circles) and in the CFO layer (green circles).
Table 1. Mean Values of the (c/a) Ratio in BTO and in CFO Extracted from the XRD Data, as a Function of the CFO Thickness for Two Different BTO Thicknesses (5 and 10 nm) CFO [nm]
2.2
(c/a)BTO (c/a)CFO
1.046 1.074
(c/a)BTO (c/a)CFO
1.043 1.074
4.4 6.6 For 5 nm BTO 1.043 1.038 1.055 1.040 For 10 nm BTO 1.039 1.032 1.044 1.032
10
15
1.029 1.021
1.015 1.000
1.018 1.008
1.005 1.000
of the 4.4 nm CFO/10 nm BTO film obtained by scanning transmission electron microscopy (STEM), together with the corresponding EDX spectroscopy images for Sr, Ti, Ba, O, Co, and Fe elements. The Ti, O, Co, and Fe chemical maps were obtained using the K edge windows, whereas the L edges were used for the Sr and Ba ones. Note that the Ba and Ti maps look very similar because of the overlapping of the K edge of Ti (4.51 keV) and L edge of Ba (4.46 keV). The spectral resolution of the equipment is here not sharp enough to separate the two contributions. The EDX analysis confirms the good chemical homogeneity of the layers and the absence of interdiffusion of the chemical species through the two interfaces. The two BTO/ STO and CFO/BTO interfaces are abrupt, the second one revealing at higher resolution (see the STEM HAADF image in Figure 2b) small height fluctuations of 1 or 2 unit cells at the surface of the BTO film. Even at small thickness, a close inspection of the BTO and CFO lattices gives evidence of dislocations in the two types of layers. The STEM HAADF image at the atomic scale has been filtered using FFT in order to show exclusively the lattice planes perpendicular to the film plane. The dislocations in BTO (at the center of the red circles) are located either close to the BTO/STO interface (at 1.7 nm from it) or at the CFO/BTO interface. They occur with an irregular spacing along the film plane; the distance between two
The c and a lattice parameters of BTO obtained in this study are comparable to those reported for thin films grown under biaxial compressive stress, the exact maximal values depending on the chosen substrate.28,29 However, all previous analyses assumed a homogeneous strain state and, consequently, a critical film thickness for relaxation, independently of the presence of an additional interface. In the same way, strain-phase diagrams were generated via thermodynamic analyses by assuming homogeneous domains.28−30 The original results of our structural analysis are (i) the clear and unexpected dependence of the BTO strain state as a function of the thickness of the CFO overlayer and (ii) the inhomogeneous strain in BTO with the presence of both relaxed and strained regions even for the thinnest CFO D
DOI: 10.1021/acsami.8b09499 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Research Article
ACS Applied Materials & Interfaces
Figure 3. (a) PFM phase images obtained from a 10 nm BTO film covered with either 5 or 15 nm of CFO, after poling the samples at ±6 V according to the schematic on the right. (b) PFM phase images obtained after ±8 V poling from a 5 nm BTO film with 5, 10, and 15 nm of CFO. (c) Sketch of the experimental XAS geometry. (d) Energy levels of the Ti4+ 3d orbitals in cubic Oh and tetragonal C4v environment, for which a positive and a negative splitting are successively considered. (e) XAS spectra at the L2,3 edge of Ti using H or V light polarization and X-ray linear dichroism (XLD) = XASH − XASV spectrum together with its simulation for 10 nm BTO (and 2.2 nm CFO).
evolves from highly strained (c/a = 1.05) to relaxed (c/a = 1.015). The fact that the BTO polarization is hardly switchable in the strained situation whereas it may be easily totally reversed in the relaxed case strongly suggests a direct correlation between the polarization magnitude P and the c/a tetragonality: the polarization P would be enhanced with increasing c/a ratio. Accordingly, because of its larger BTO c/a ratio (1.015 with respect to 1.005), the phase contrast for the 15 nm CFO/5 nm BTO film is found higher than for the 15 nm CFO/10 nm BTO film. To summarize, the more tetragonal is the BTO layer, the more stable and likely higher is the polarization and the harder is its switchability. The PFM analysis also reveals that the CFO overlayer is stabilizing large μm-wide BTO domains with a single FE polarization orientation, pointing outward. This stabilization effect of CFO on the polarization of BTO has recently been reported for heterostructures grown on silicon.16 Large coriented polarized domains had similarly been observed with Co/BTO layers.36 The polarization state of the BTO strained layers has been further analyzed through the X-ray linear dichroism (XLD) of the Ti atoms.37 Indeed, the electric polarization of BTO at 300 K is due to the collective displacement of the positive ions relative to the negative ones along the [001] direction, the higher displacement concerning the Ti4+ (3d0 4s0) ions. An asymmetry in the Ti electronic charge distribution is therefore expected, which may be revealed by differences in the X-ray absorption spectra of Ti at the L2,3 edge (2p6 3d0 → 2p5 3d1), when using either a horizontal (H) or a vertical (V) light polarization direction. The experimental geometry is sketched in Figure 3c. The X-ray wave vector is at 30° of the surface, a geometry generally reported as grazing with respect to the film plane. In the H polarization, the electric field vector may be considered as parallel to the c-axis (also denoted z-axis), whereas in the V polarization, it may be considered as parallel to the a-axis (also denoted y-axis). As a consequence, the X-ray absorption with the H (V) polarization is expected to concern predominantly orbitals with z(x, y) character. In a cubic perovskite structure,
dislocations may vary between 50 and 14 nm close to the STO substrate while it may be of 5 nm close to CFO. Because of the CFO overlayer, the relaxation of BTO occurs preferentially in its upper layers while an enhanced tetragonality is established in the lower layers. The dislocations in CFO (small green circles) are much more numerous than in BTO; their spacing may be lower than 5 nm. This is in agreement with a relaxation of the CFO overlayer starting before the bottom BTO layer relaxes itself. 3.2. Tetragonality and Ferroelectricity of BTO. An enhanced electric polarization has been associated to highly strained BTO thin films.28,35 The FE nature of our films was studied by PFM at room temperature. The bias voltage was applied to the tip while the Nb-doped STO substrate was grounded. Figure 3a shows two typical phase images collected after writing different rectangle areas at ±6 V on a 5/10 nm and on a 15/10 nm CFO/BTO film, respectively. The written zone covers a 4 × 5.4 μm2 wide area, over a 7 × 7 μm2 investigated surface. The two PFM phase images are well-contrasted and it could be checked that the contrast persists over several weeks. The first observation concerns the orientation of the FE polarization in the BTO layer. Because the phase does not change after applying a negative voltage, we can conclude that the polarization in the as-grown films is pointing along the normal to the film, from the BTO/STO interface toward the CFO/BTO one. The second observation concerns the phase contrast, which is found higher when the CFO overlayer is thicker. Knowing that the BTO film progressively relaxes under the influence of the CFO overlayer, we may infer that the more the BTO layer is strained with respect to the substrate, the more the polarization pointing outward is stable and the harder is its reversal. This behavior is confirmed considering a thinner 5 nm BTO film. Figure 3b accounts for the different BTO-related PFM responses, at given writing voltage (±8 V), when the CFO overlayer thickness increases from 5 to 15 nm: a maximum phase contrast of 180° could be reached with the thicker CFO film whereas no contrast could be detected with the thinner CFO film. The XRD analysis has shown that the 5 nm BTO layer E
DOI: 10.1021/acsami.8b09499 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Research Article
ACS Applied Materials & Interfaces
Figure 4. (a) Energy levels of the 3d orbitals for Co2+ in an octahedral site and for Fe3+ in octahedral or tetrahedral sites of tetragonal CFO, with related atomic geometries. (b) XLD (i.e. XASH-XASV) spectra at the L2,3 Co edge. (c) XLD spectra at the L2,3 Fe edge.
two peaks related to the z2 and x2 − y2 orbitals were measured with opposite signs. In these reports, the BTO lattice had a small c/a ratio (bulk or close to the bulk value). Our BTO films display at low thickness a very large c/a ratio (>1) and the splitting of the eg subgroup appears as opposite: the z2 orbital shifts below x2 − y2. Such inversion of the splitting has already been observed in the context of LaAlO3/STO interfaces and similarly explained on the basis of elongated octahedrons.39,40 3.3. Tetragonality of CFO. The CFO oxide consists in Co2+ ions occupying preferentially the octahedral sites of the face centered cubic O2− sublattice and in Fe3+ ions distributed in nearly equal proportion over octahedral and tetrahedral sites. The ideal inverse spinel formula unit (f. u.) writes Fet3+(Co2+Fe3+)oO42−, where the label (t) refers to tetrahedral sites and the label (o) refers to octahedral sites. We have demonstrated the marked inverse spinel character (80−90%) of our AO-MBE grown CFO layers in a previous report.27 We investigate herein the specific tetragonal character of the thinnest CFO overlayers by XLD at the L2,3 edge of Co and Fe. In a FM film, the linear dichroism may arise from two effects: one related to magnetization M, the signal being proportional to ⟨M⟩2, the other one related to an asymmetric orbital contribution.41 In the following, we focus the analysis on the XLD features which may be assigned to the asymmetry induced in the electronic configuration of Co2+ and Fe3+ by the tetragonality. Figure 4a depicts the changes in the (oneelectron-like) energy levels of the Co2+ ion (3d7 4s0), associated to an expansion of the unit cell of CFO along the c axis. The distorted octahedral environment of Co2+ by the O2− ions caused by the in-plane compressive stress of the substrate (c/a > 1) should lead to the Jahn−Teller effect, that is, the asymmetric occupation of the t2g orbitals: filled dxz and dyz with respect to half-filled dxy. For Fe3+ (3d5 4s0), every d orbital is singly occupied but the lattice expansion along c should also lead to the splitting of each t2g and eg subgroup. The expected energy distributions of the d levels for Fe3+ in octahedral and in tetrahedral site for elongated unit cells (c/a > 1) are sketched in Figure 4a. Figure 4b,c shows the XLD (XASH-XASV) spectra for the two types of atoms (Co and Fe) in CFO considering a fixed BTO thickness (10 nm) and three different CFO thicknesses (2.2, 4.4,
the Ti4+ ion is at the center of the octahedron formed by the O2− ions and the empty 3d orbitals are split in two groups (t2g and eg) according to the Oh symmetry. In the tetragonal structure, the displacement of the Ti4+ ion along z will lead to a new symmetry and hence to a new energy repartition of the d orbitals.37 The hierarchy of the energy levels is shown in Figure 3d considering successively the Ti4+ ion at the center of a cubic octahedron, the Ti4+ ion displaced along z from this center, and finally, the Ti4+ ion displaced along z from the center of a markedly distorted (expanded along z) octahedron. Figure 3e shows the XAS spectra and the resulting XLD Ti spectrum (XASH-XASV) for the 2.2 nm CFO/10 nm BTO film. Although the XAS signal magnitude decreases rapidly with the thickness of the CFO layer, because of the limited mean free path of the electrons, we could check the good reproducibility of this XLD Ti signal by considering several tetragonal CFO/BTO films. The high level of tetragonality in our ultrathin AO-MBE grown BTO layers allows us to extract with confidence the XLD Ti characteristics of c-axis polarized BTO domains. The Ti L3 dichroic features consist first in a very faint positive peak at 458.4 eV, followed by a distinct negative peak at 458.7 eV, and then by broader positive and negative contributions on either side of 460.75 eV. The same is reproduced with different widths and intensities at the L2 edge. The t2g and eg energy levels are known to be split by about 2 eV. Qualitatively, the two strongest XLD peaks at 460.4 and 461.0 eV may thus be related to the nondegenerate eg orbitals, that is, to z2 and x2 − y2 orbitals, while the peak at 458.7 may be assigned to the xy orbitals. The fact that the xz and yz orbitals do not lead to a clear-cut positive signal could be due to a partial occupation of these orbitals through hybridization or to a mixing of the excited energy states. Ligand field multiplet calculations were performed to simulate the XLD spectrum. The best fit was obtained for the crystal field parameters 10Dq = 1.8 eV, Ds = 0.04 eV, and Dt = 0.04 eV, where 10Dq denotes the energy splitting between the t2g and eg states in the cubic Oh symmetry, whereas Ds and Dt take into account the tetragonal distortion and the Ti4+ displacement (C4v symmetry). The positive values of Ds and Dt induce a shift of the z2 level below the x2 − y2 one, as depicted in Figure 3d. Our XLD data may be compared to those reported for c-axis domains in bulk BTO38 or in CFO/BTO nanocomposite films,37 where the F
DOI: 10.1021/acsami.8b09499 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Research Article
ACS Applied Materials & Interfaces
Figure 5. (a) Distribution and related magnetic moments of Fe3+ and Co2+ ions in tetrahedral (t) or octahedral (o) sites of CFO. (b) XAS spectrum at the L2,3 edge of Co using right (R) or left (L) light polarization, the related XAS average and the X-ray magnetic circular dichroism (XMCD) spectrum, for 15 nm CFO/10 nm BTO. A magnetic field of 5 T was applied in-plane. (c) Same at the L2,3 edge of Fe.
different CFO/BTO films on the basis of XAS magnetic circular dichroism measurements. The geometry of the XAS experiment using left (L) or right (R) circular polarized light is shown in Figure 5b. The X-ray magnetic circular dichroism (XMCD) spectrum is defined as (XASR-XASL). The XAS and XMCD spectra, measured at the L2,3 edges of Co and Fe on the thickest 15 nm CFO/10 nm BTO film, are shown in Figure 5b,c. A magnetic field of 5 T was applied along the light wave vector direction, at 30° with respect to the sample surface. The field of 5 T being fairly larger than the coercitive field of CFO, the magnetization is at its maximum as well as the dichroic signal, which is reflecting the average magnetic moment of each atomic species. The different Fe and Co XMCD spectral features may be further interpreted on the basis of ligand field multiplet calculations or on the basis of the configuration-interaction (CI) model, after taking into account the 3d electron configuration of the cation, the crystal field for each type of site, the spin−orbit coupling, the electron−electron interaction, and the hybridization effects.42−44 Whereas the simulation of the Co XMCD spectra still remains challenging,21 the Fe XMCD spectra in spinels are well-understood and reproduced by the calculations.45,46 As observed in Figure 5c for the 15 nm thick CFO overlayer, the dichroic signal related to Fe has a strong positive component at 708.6 eV which is solely due to the Fe3+ ions in the tetrahedral sites (Fet3+) and a strong negative component at 709.3 eV which is solely due to the Fe3+ ions in the octahedral sites (Feo3+). The small negative component at 707.4 eV is also related to (Feo3+). Distinctly different, the dichroic signal of Co in Figure 5b is essentially negative. It comes from the Co2+ ions in the octahedral sites (Coo2+). In a first step, we analyzed the Co XMCD spectra associated to the different CFO/BTO films on the basis of the so-called sum rules (SRs).47,48 Taking the notation of Chen et al.,47 we considered the simplified formulas (in μB units)
and 6.6 nm). The results were found similar for the 5 nm thick BTO films. As a whole, the intensity of the Co XLD peaks increases with increasing CFO thickness. Focusing on the shape of the Co XLD signal, we observe that the thinnest CFO deposits (2.2 and 4.4 nm) depart from the 6.6 nm deposit essentially by a marked inflexion of the signal at 776.1 eV (Figure 4b). Assuming the hierarchy of the Co empty 3d levels as in Figure 4a, this inflexion could be due to the shift of xyo (z2o) toward higher (lower) energies when the CFO thickness decreases. Such shifts are expected from a contraction (expansion) of the distance between the Co2+ ion and its surrounding O2− ions along the a (c) axis, in other words from an increase of the tetragonality (c/a ratio). The changes in the Fe XLD spectrum from 4.4 to 2.2 nm, at 707.6 and 709.2 eV, could similarly to the Co case be provoked by the enhanced tetragonality of the thinnest CFO deposit. If we consider by order of increasing energy, the (xzoyzo), (x2 − y2t), (z2t), (xyo), (z2o), and (xyt) levels for Fe3+, the increase of c/a should induce a shift of the (z2o) and (xyt) levels toward lower energies. Although the assignment of the Fe XLD peaks to specific d orbitals has speculative character, such shifts would explain the two strongest changes observed in the Fe XLD spectrum (Figure 4c). The electronic levels of the two types of ions are thus subject to measurable XLD changes under the effect of the tetragonal distortion of the CFO overlayer. To our knowledge, this is the first experimental report on this topic. These changes in the electronic structure of the ions in tetragonal CFO will have consequences on their magnetic moments. 3.4. Magnetic Long-Range Ordering of CFO. In the ideal Fet3+(Co2+Fe3+)oO2−4 inverse spinel, the average magnetic moment of the Fe3+ cations is close to zero because of the prevailing antiferromagnetic coupling between octahedral (Oh) and tetrahedral (Td) sites (see Figure 5a). Therefore, the magnetization in CFO comes essentially from the Co2+ cations, and the magnetic moment per CFO f. u. should be close to the magnetic moment of one Co2+ cation. If we assume the quenching of the orbital moment (ml) of the cations, as usually considered in most of oxides, the total magnetic moment (m) of each ion is equal to its spin moment (ms), that is, about 3 μB for Co2+ and 5 μB for Fe3+. At saturation, a magnetic moment of 3 μB per CFO f. u. is thus expected. In the following, we analyze in detail the magnetic moments of the Co and Fe ions in the
ml = −2qNh /(3r )
(1)
ms = −(3p − 2q)Nh /r
(2)
where Nh is the number of 3d holes, p is the integral of the XMCD signal over the L3 edge, q is the integral of the XMCD signal over the L3 and L2 edges, and r is the integral of the average G
DOI: 10.1021/acsami.8b09499 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Research Article
ACS Applied Materials & Interfaces
Figure 6. (a) Integrals of the XMCD spectra at the L2,3 edge of Co for different CFO/BTO films (fixed BTO thickness). (b) SRs-derived (ml/ms) ratio of Co, where ml is the average orbital magnetic moment and ms is the average spin magnetic moment, as a function of the CFO thickness for two BTO thicknesses. (c) Relative evolution of the average ml, ms, and m = ml + ms magnetic moments of the Co ion deduced from the SRs, as a function of the CFO thickness.
Figure 7. (a) Experimental XMCD spectrum at the L2,3 edge of Fe for 2.2 nm CFO on 5 nm BTO and its simulation. (b) Same for 15 nm CFO on 5 nm BTO. (c) Respective percentages of the Fe2+ ion in octahedral site, of the Fe3+ ion in tetrahedral site, and of the Fe3+ ion in octahedral site, deduced from multiplet simulation. (d) Calculated average magnetic moments of the Fe ion and of the Co ion as a function of the CFO thickness for two BTO thicknesses, using the Fe percentages of Figure 7c and α = 1(6) for 10(5) nm BTO.
be determined with good accuracy from the XMCD integral spectra alone. The integral signals of the Co XMCD spectra from different CFO overlayers on 10 nm BTO are displayed in Figure 6a. The derived (ml/ms) values for Co are displayed in Figure 6b, together with those related to the 5 nm BTO layer.
XAS spectrum after subtraction of the step background signal between L3 and L2. The SRs allow evaluating on the one hand the (ml/ms) ratio and on the other hand the average total magnetic moment m = ml + ms of the Co ion. The ratio (ml/ms) = (2/3q)/(3p − 2q) only relies on the parameters p and q. It may H
DOI: 10.1021/acsami.8b09499 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Research Article
ACS Applied Materials & Interfaces
thickness decreases. The most significant raise is observed with the 5 nm BTO bottom layer, the (Feo2+) percentage reaching 23.5% for 2.2 nm CFO thickness, whereas with 10 nm BTO it keeps relatively low (7.5%). The thinnest heterostructure is thus subject to the strongest chemical alteration. However, for 4.4 nm CFO thickness, the fraction of (Feo2+) ions is already divided by a factor 3.5. On the basis of this significant decrease of the (Feo2+) percentage from 2.2 to 4.4 nm CFO thickness, one may infer that the chemically modified CFO region with (Feo2+) ions is confined to a 2.2 nm thickness, starting from the atomic CFO/ BTO interface layer. The second observation is that the (Fet3+) proportion, that is, the inversion degree of the spinel, decreases with increasing (Feo2+) proportion, or equivalently with decreasing CFO thickness. Remarkably, we have noticed that the fraction in CFO of (Feo3+) or (Fet3+) with respect to (Fe3+) is independent of the BTO thickness. Our results agree very well with those of recent reports on epitaxial CFO layers deposited on Al2O3/Si(111)21 or on BTO(001) substrate.50 In the first case, the (Feo2+) ion in a proportion of 20% was similarly detected in an ultrathin CFO deposit (1.4 nm).21 In the second case, the dominating presence of the (Feo2+) ion was revealed at 2 nm from the interface, the CFO recovering its ideal composition at 7 nm.50 Lastly, we used the (Fet3+), (Feo3+), and (Feo2+) proportions deduced from the Fe XMCD fits to calculate the expected average magnetic moment of Fe. In accordance with the ligandfield multiplet calculations, the magnetic moment of the Fe3+ ion was set equal to m = ms = 5 μB and the magnetic moment of the Fe2+ ion was set equal to m = ms + ml = 4.23 μB. As displayed in Figure 7d, the calculated average magnetic moment of Fe (⟨mFe⟩calc) increases with decreasing CFO thickness, mirror-like the average magnetic moment of Co (⟨mCo⟩SR) shown in Figure 6c, which was deduced independently from the SRs. These opposite behaviors of ⟨mFe⟩calc and ⟨mCo⟩SR lead to conclude that the Co2+ ion concentration decreases concomitantly to the formation of the Fe2+ ion. 3.5. Oxidoreduction of CFO at the Interface with BTO. Two mechanisms may a priori be invoked concerning the presence of the Fe2+ ions in CFO near the interface with BTO (in a 0−2.2 nm thick interfacial region), while preserving the electroneutrality of the layers: the reduction of Fe3+ into Fe2+ ions either (i) with the simultaneous formation of oxygen vacancies or (ii) with the simultaneous oxidation of Co2+ into Co3+. The second mechanism can, alone, account for the drastic decrease of ⟨mCo⟩SR when decreasing the CFO thickness below 5 nm, but the occurrence of the first mechanism cannot be totally ruled out. Envisaging both mechanisms as probable, we have considered the general formula for the CFO spinel:
First, the (ml/ms) ratio of the Co ions in the thickest CFO deposit has a noticeable value of about 0.36. This value is very close to that reported by Wakabayashi et al.,21 who simulated the Co XMCD spectra by the CI cluster model. Hence, the quenching of the orbital moment of the Co2+ ions in cubic CFO is far from being total. Secondly, the (ml/ms) ratio increases with decreasing CFO thickness up to ∼0.55. One also notice that this (ml/ms) increase is more important with the thicker BTO film, as is the CFO (c/a) increase (see Table 1). In other words, the (ml/ms) ratio seems directly related to the (c/a) tetragonality. The XMCD analysis confirms the XLD results concerning the influence of strain on the electronic configuration of the ions and as a consequence on their magnetic moment. The integrals of the Co average XAS spectra were thereafter considered to extract the r values and finally, from eqs 1 and 2, the average ⟨ml⟩ and ⟨ms⟩ values of the Co ion. By taking Nh = 3, the derived ⟨mCo⟩SR value for the thickest film was found equal to 1.4 μB. This relatively small value for Co2+ can be explained by a combination of factors: the lattice distortion (dislocations), the reduced dimension (thin film configuration), the presence of a fraction of direct spinel in the film, weakened antiferromagnetic couplings, and eventually a nonperfectly adapted usage of the SR. Nevertheless, within a very same data treatment, the trends observed for ⟨mCo⟩SR when decreasing the CFO thickness are believed to be meaningful. Figure 6c displays the evolution of the SR-derived ⟨ml⟩, ⟨ms⟩, and ⟨m⟩ values of the Co ion as a function of the CFO thickness. From 15 to 5 nm, the average magnetic moment of the Co ion slowly decreases while below 4.4 nm, it drastically drops. This abrupt decrease cannot be explained by a sudden crystalline deterioration or a concomitant sudden increase of antiphase boundaries. Changes in the ion nature and/or in the ion repartition near the interface with BTO may therein be inferred. This will be confirmed in the second step of the analysis dedicated to the Fe XMCD spectra. Figure 7a,b shows the Fe XMCD spectra of the 2.2 nm CFO/ 5 nm BTO film and of the 15 nm CFO/5 nm BTO film, respectively. The spectrum in Figure 7b is very close to that of Figure 5c and consistent with a quasi-inverse spinel. The spectrum in Figure 7a is radically different. The positive feature at 708.6 eV due to the (Fet3+) ions has almost vanished while a very strong negative signal is observed at 707.4 eV. This negative signal, together with the very faint negative additional contribution at 705.3 eV, can be related to the presence of a large fraction of Fe2+ ions in octahedral sites. The proportion of Fe2+ with respect to Fe3+ in CFO is generally lower than 10%;27,46 this is actually the case for the 15 nm CFO film, but not for the 2.2 nm CFO film. To determine quantitatively the proportions of the different Fe ions in our films, we have used theoretical spectra specific to the (Fet3+), (Feo3+), and (Feo2+) ions in ferrite spinels, calculated within the multiplet approach.49 It may be noted that in this calculation the quenching of the orbital moment of the Fe3+ ion is found total (m = ms), whereas the (ml/ms) ratio for the Fe2+ ion is found equal to 0.22. The experimental Fe XMCD spectra of the different films were fitted by least-squares refinement, using a linear combination of the theoretical spectra. The only free parameters in the fits were the respective contributions of each type of cation, that is, the respective percentages of (Fet3+), (Feo3+), and (Feo2+) ions. Two representative fits are shown in Figure 7a,b. The evolution of the different ion percentages as a function of the CFO thickness for two fixed BTO thicknesses is displayed in Figure 7c. Whatever the BTO thickness, the (Feo2+) ion percentage, which is of only 2−3% for 15 nm CFO thickness, increases when the CFO
[Fey 3 + Co2 +1 − y]t (Co3 + x / α Co2 + y − x / α Fe3 +2 − y − xFe 2 + x )o O2 −4 − x /2 + x /(2α)
(3)
where y is the inversion parameter, x (x/α) is the fraction of Fe2+(Co3+) ions in octahedral sites, and α being the ratio Fe2+/ Co3+. We have calculated the average magnetic moment of Co (⟨mCo⟩calc) according to formula 3, using the x, y values deduced from the Fe proportions of Figure 7c and different α. We have assumed a low spin configuration for Co3+,51 that is, a magnetic moment equal to zero, and considered for Co2+ a magnetic moment equal to m = ms + ml = 2.6 + 1.0 = 3.6 μB.21 The assumption of a prevailing antiferromagnetic coupling between (o) and (t) sites has been preserved. The Co average magnetic moment (⟨mCo⟩calc), calculated with α = 1 for 10 nm BTO and α I
DOI: 10.1021/acsami.8b09499 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Research Article
ACS Applied Materials & Interfaces
correlation between strain, ferroelectricity, and ferromagnetism in layered artificial multiferroic systems and give highlights on the important chemical/electronic changes arising at interfaces. In the context of CFO-based magnetic tunnel junctions or BTObased electric tunnel junctions, the knowledge of such interface effects is essential.
= 6 for 5 nm BTO, is plotted in Figure 7d. The agreement with the ⟨mCo⟩SR evolutions of Figure 6c is reasonably good. Notice that for the 2.2(15) nm CFO layer, the α = 6 value for 5 nm BTO results in a percentage of Co2+ changed into Co3+ equal to 8(1) %, while for 10 nm BTO the α = 1 value results in a fraction equal to 15(4) %. For example, with the 10 nm thick BTO layer, the nominal composition of CFO evolves from [Fe3+0.73Co2+0.27]t (Co 3+ 0.15 Co 2+ 0. 58 Fe 3+ 1.12 Fe 2+ 0.15 ) o O 2− 4 at 2.2 nm to [Fe3+0.88Co2+0.12]t (Co3+0.04Co2+0.84Fe3+1.08Fe2+0.04)oO2−4 at 15 nm. With the 5 nm thick BTO layer, the Co3+ percentages in CFO are lower because of the presence of oxygen vacancies, but still significant at 2.2 nm. In our model, whatever the CFO thickness, the average magnetic moment per f.u. keeps close to 3.7 μB. For the ultrathin CFO layers of 2.2 nm, and thus close to the interface with BTO, the drastic decrease of ⟨mCo⟩ is compensated by the increase of ⟨mFe⟩. The FM ordering is thus preserved in ultrathin CFO layers down to 2 nm, although important chemical changes occur close to the BTO interface. We propose two explanations for the oxidoreduction process at the interface with BTO. The very strained tetragonal state of the CFO overlayer in the ultrathin thickness range by decreasing the distance between the Co2+ and Fe3+ ions could lead to the reduction of some of the Fe3+ ions into Fe2+ and the concomitant oxidation of Co2+ ions into Co3+. Another contributing factor could be the stabilization of the BTO polarization by charges provided by the CFO overlayer. We have shown that the cpolarized domains of the strained BTO bottom layers, with polarization pointing outward, are stabilized by the top CFO layer. This stabilization by CFO implies an imperfectly insulating character of the CFO layer, that is, the accumulation of negative charges toward the interface with BTO in order to compensate the polarization (or at least the polarity of BTO). This could explain why Fe2+-rich regions are formed in CFO close to the BTO interface. As we do not have access to the concentration profile of each type of ion in the CFO layers with sub-nanometer resolution, we cannot draw a definitive conclusion on the exact mechanisms at the origin of the oxidoreduction process.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. ORCID
Nathalie Jedrecy: 0000-0001-7786-8658 Author Contributions
The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS The authors acknowledge the “Agence Nationale de la Recherche (ANR)” for their funding through the IOBTO project (grant no. ANR-15-CE09-0005-01). This work was also supported in part by “Triangle de la Physique” and “Ile-deFrance” (C’Nano IdF, DIM-OXYMORE and ISC-PIF) under the IMAFMP and MAEBA grants. The authors acknowledge the SOLEIL Synchrotron for supplying beamtime and the technical staff of the DiffAbs, SixS, and DEIMOS beamlines for assistance during the experiments. T.A. and D.S. would like to express their thanks to Cindy L. Rountree from CEA-SPEC for the PFM experiments. X.P. from CIMAP laboratory would like to acknowledge the ANR for their funding through an EQUIPEX project (grant no. ANR-11-EQPX-0020) and also the Normandie Region for its financial support that allowed the use of the JEOL ARM200F microscope and of the FEI FIB system. N.J. and X.P. acknowledge the financial support from the CNRS-CEA “METSA” French network (FR CNRS 3507) for the TEM experiments conducted on the IRMA platform.
■
4. CONCLUSIONS We have investigated a series of CFO/BTO heterostructures with different BTO and CFO individual thicknesses. In this way, we gave evidence that the BTO FE and the CFO FM responses are both dependent on the strain state of the full stack, which ranges from highly tetragonal to nearly cubic. Importantly, we highlight the fact that the strain release in a bilayer system has much more complexity and difference than that in a single layer because a cross-correlation exists between the two types of layers. Via strain, the two BTO/substrate and CFO/BTO interfaces play a critical role in the final structural and physical properties. The magnitude and switchability of the (out-ofplane) electric polarization in the BTO layer depend not only on the (c/a) ratio reached in the strained tetragonal state but also on the thickness of the top CFO layer. Lastly, we used XLD and XMCD to get a comprehensive survey of the nature and distribution of the cations in the two types of layers and of the magnetic moments in CFO. If the magnitude of the saturation magnetization in CFO may be preserved for very thin CFO layers down to 4.4 nm and even below, important chemical changes (oxidoreduction) occur close to the BTO interface, probably driven by polarity compensation, and which may have great impact on the magneto-electric response of the multiferroic heterostructure. All our results stress the complex cross-
REFERENCES
(1) Khomskii, D. I.; Sawatzky, G. A. Interplay between Spin, Charge and Orbital Degrees of Freedom in Magnetic Oxides. Solid State Commun. 1997, 102, 87−99. (2) Takagi, H.; Hwang, H. Y. An Emergent Change of Phase for Electronics. Science 2010, 327, 1601−1602. (3) Bibes, M.; Villegas, J. E.; Barthélémy, A. Ultrathin Oxide Films and Interfaces for Electronics and Spintronics. Adv. Phys. 2011, 60, 5−84. (4) Ha, S. D.; Ramanathan, S. Adaptive Oxide Electronics: A Review. J. Appl. Phys. 2011, 110, 071101. (5) Cheong, S.-W.; Mostovoy, M. Multiferroics: a Magnetic Twist for Ferroelectricity. Nat. Mater. 2007, 6, 13−20. (6) Ramesh, R.; Spaldin, N. A. Multiferroics: Progress and Prospects in Thin Films. Nat. Mater. 2007, 6, 21−29. (7) Heron, J. T.; Schlom, D. G.; Ramesh, R. Electric Field Control of Magnetism using BiFeO3-based heterostructures. Appl. Phys. Rev. 2014, 1, 021303. (8) Choi, E.; Fix, T.; Kursumovic, A.; Kinane, C. J.; Arena, D.; Sahonta, S.-L.; Bi, Z.; Xiong, J.; Yan, L.; Lee, J.-S.; Wang, H.; Langridge, S.; Kim, Y.-M.; Borisevich, A. Y.; MacLaren, I.; Ramasse, Q. M.; Blamire, M. G.; Jia, Q.; MacManus-Driscoll, J. L. Room Temperature Ferrimagnetism and Ferroelectricity in Strained, Thin Films of BiFe0.5Mn0.5O3. Adv. Funct. Mater. 2014, 24, 7478−7487. (9) Lu, C.; Hu, W.; Tian, Y.; Wu, T. Multiferroic Oxide Thin Films and Heterostructures. Appl. Phys. Rev. 2015, 2, 021304. J
DOI: 10.1021/acsami.8b09499 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Research Article
ACS Applied Materials & Interfaces
Grown by Atomic Oxygen Plasma Assisted Molecular Beam Epitaxy. J. Appl. Phys. 2012, 112, 114116. (27) Aghavnian, T.; Moussy, J.-B.; Stanescu, D.; Belkhou, R.; Jedrecy, N.; Magnan, H.; Ohresser, P.; Arrio, M.-A.; Sainctavit, P.; Barbier, A. Determination of the Cation Site Distribution of the Spinel in Multiferroic CoFe2O4/BaTiO3 Layers by X-ray Photoelectron Spectroscopy. J. Electron Spectrosc. Relat. Phenom. 2015, 202, 16−21. (28) Choi, K. J.; Biegalski, M.; Li, Y. L.; Sharan, A.; Schubert, J.; Uecker, R.; Reiche, P.; Chen, Y. B.; Pan, X. Q.; Gopalan, V.; Chen, L.Q.; Schlom, D. G.; Eom, C. B. Enhancement of Ferroelectricity in Strained BaTiO Thin Films. Science 2004, 306, 1005−1009. (29) Schlom, D. G.; Chen, L.-Q.; Eom, C.-B.; Rabe, K. M.; Streiffer, S. K.; Triscone, J.-M. Strain Tuning of Ferroelectric Thin Films. Annu. Rev. Mater. Res. 2007, 37, 589−626. (30) Li, Y. L.; Chen, L. Q. Temperature-Strain Phase Diagram for BaTiO3 Thin Films. Appl. Phys. Lett. 2006, 88, 072905. (31) Li, C.; Huang, L.; Li, T.; Lü, W.; Qiu, X.; Huang, Z.; Liu, Z.; Zeng, S.; Guo, R.; Zhao, Y.; Zeng, K.; Coey, M.; Chen, J. S.; Ariando; Venkatesan, T. Ultrathin BaTiO3-Based Ferroelectric Tunnel Junctions through Interface Engineering. Nano Lett. 2015, 15, 2568−2573. (32) Gao, X. S.; Bao, D. H.; Birajdar, B.; Habisreuther, T.; Mattheis, R.; Schubert, M. A.; Alexe, M.; Hesse, D. Switching of Magnetic Anisotropy in Epitaxial CoFe2O4 Thin Films Induced by SrRuO3 Buffer Layer. J. Phys. D: Appl. Phys. 2009, 42, 175006. (33) Gutiérrez, D.; Foerster, M.; Fina, I.; Fritsch, D.; Fontcuberta, J. Dielectric Response of Epitaxially Strained CoFe2O4 Spinel Thin Films. Phys. Rev. B: Condens. Matter Mater. Phys. 2012, 86, 125309. (34) Zhang, Y.; Shen, L.; Liu, M.; Li, X.; Lu, X.; Lu, L.; Ma, C.; You, C.; Chen, A.; Huang, C.; Chen, L.; Alexe, M.; Jia, C.-L. Flexible QuasiTwo-Dimensional CoFe2O4 Epitaxial Thin Films for Continuous Strain Tuning of Magnetic Properties. ACS Nano 2017, 11, 8002−8009. (35) Garcia, V.; Fusil, S.; Bouzehouane, K.; Enouz-Vedrenne, S.; Mathur, N. D.; Barthélémy, A.; Bibes, M. Giant Tunnel Electroresistance for Non-destructive Readout of Ferroelectric States. Nature 2009, 460, 81−84. (36) Jedrecy, N.; von Bardeleben, H. J.; Badjeck, V.; Demaille, D.; Stanescu, D.; Magnan, H.; Barbier, A. Strong Magnetoelectric Coupling in Multiferroic Co/BaTiO3 Thin Films. Phys. Rev. B: Condens. Matter Mater. Phys. 2013, 88, No. 121409(R). (37) Schmitz-Antoniak, C.; Schmitz, D.; Borisov, P.; de Groot, F. M. F.; Stienen, S.; Warland, A.; Krumme, B.; Feyerherm, R.; Dudzik, E.; Kleemann, W.; Wende, H. Electric In-plane Polarization in Multiferroic CoFe2O4/BaTiO3 Nanocomposite Tuned by Magnetic Fields. Nat. Commun. 2013, 4, 2051. (38) Chopdekar, R. V.; Malik, V. K.; Fraile Rodríguez, A.; Le Guyader, L.; Takamura, Y.; Scholl, A.; Stender, D.; Schneider, C. W.; Bernhard, C.; Nolting, F.; Heyderman, L. J. Spatially Resolved Strain-Imprinted Magnetic States in an Artificial Multiferroic. Phys. Rev. B: Condens. Matter Mater. Phys. 2012, 86, 014408. (39) Salluzzo, M.; Gariglio, S.; Torrelles, X.; Ristic, Z.; Di Capua, R.; Drnec, J.; Sala, M. M.; Ghiringhelli, G.; Felici, R.; Brookes, N. B. Structural and Electronic Reconstructions at the LaAlO3/SrTiO3 Interface. Adv. Mater. 2013, 25, 2333−2338. (40) Lesne, E.; Reyren, N.; Doennig, D.; Mattana, R.; Jaffrès, H.; Cros, V.; Petroff, F.; Choueikani, F.; Ohresser, P.; Pentcheva, R.; Barthélémy, A.; Bibes, M. Suppression of the Critical Thickness Threshold for Conductivity at the LaAlO3/SrTiO3 Interface. Nat. Commun. 2014, 5, 4291. (41) van der Laan, G.; Arenholz, E.; Chopdekar, R. V.; Suzuki, Y. Influence of Crystal Field on Anisotropic X-ray Magnetic Linear Dichroism at the Co2+ L2,3 Edges. Phys. Rev. B: Condens. Matter Mater. Phys. 2008, 77, 064407. (42) van der Laan, G.; Figueroa, A. I. X-ray Magnetic Circular DichroismA Versatile Tool to Study Magnetism. Coord. Chem. Rev. 2014, 277−278, 95−129. (43) Ikeno, H.; de Groot, F. M. F.; Stavitski, E.; Tanaka, I. Multiplet Calculations of L2,3 X-ray Absorption Near-Edge Structures for 3d Transition-Metal Compounds. J. Phys.: Condens. Matter 2009, 21, 104208.
(10) Ortega, N.; Kumar, A.; Scott, J. F.; Katiyar, R. S. Multifunctional Magnetoelectric Materials for Device Applications. J. Phys.: Condens. Matter 2015, 27, 504002. (11) Bibes, M.; Barthélémy, A. Multiferroics: towards a Magnetoelectric Memory. Nat. Mater. 2008, 7, 425−426. (12) Zheng, H.; Wang, J.; Lofland, S. E.; Ma, Z.; Mohaddes-Ardabili, L.; Zhao, T.; Salamanca-Riba, L.; Shinde, S. R.; Ogale, S. B.; Bai, F.; Viehland, D.; Jia, Y.; Schlom, D. G.; Wuttig, M.; Roytburd, A.; Ramesh, R. Multiferroic BaTiO3-CoFe2O4 Nanostructures. Science 2004, 303, 661−663. (13) Chu, Y.-H.; Martin, L. W.; Holcomb, M. B.; Gajek, M.; Han, S.-J.; He, Q.; Balke, N.; Yang, C.-H.; Lee, D.; Hu, W.; Zhan, Q.; Yang, P.-L.; Fraile-Rodríguez, A.; Scholl, A.; Wang, S. X.; Ramesh, R. Electric-Field Control of Local Ferromagnetism using a Magnetoelectric Multiferroic. Nat. Mater. 2008, 7, 478−482. (14) Liu, M.; Obi, O.; Lou, J.; Chen, Y.; Cai, Z.; Stoute, S.; Espanol, M.; Lew, M.; Situ, X.; Ziemer, K. S.; Harris, V. G.; Sun, N. X. Giant Electric Field Tuning of Magnetic Properties in Multiferroic Ferrite/ Ferroelectric Heterostructures. Adv. Funct. Mater. 2009, 19, 1826− 1831. (15) Chen, Y.-J.; Hsieh, Y.-H.; Liao, S.-C.; Hu, Z.; Huang, M.-J.; Kuo, W.-C.; Chin, Y.-Y.; Uen, T.-M.; Juang, J.-Y.; Lai, C.-H.; Lin, H.-J.; Chen, C.-T.; Chu, Y.-H. Strong Magnetic Enhancement in Self-Assembled Multiferroic-Ferrimagnetic Nanostructures. Nanoscale 2013, 5, 4449− 4453. (16) Scigaj, M.; Dix, N.; Gázquez, J.; Varela, M.; Fina, I.; Domingo, N.; Herranz, G.; Skumryev, V.; Fontcuberta, J.; Sánchez, F. Monolithic Integration of Room-Temperature Multifunctional BaTiO3-CoFe2O4 Epitaxial Heterostructures on Si(001). Sci. Rep. 2016, 6, 31870. (17) Liu, G.; Nan, C.-W.; Xu, Z. K.; Chen, H. Coupling Interaction in Multiferroic BaTiO3−CoFe2O4 Nanostructures. J. Phys. D: Appl. Phys. 2005, 38, 2321−2326. (18) Fina, I.; Dix, N.; Rebled, J. M.; Gemeiner, P.; Martí, X.; Peiró, F.; Dkhil, B.; Sánchez, F.; Fàbrega, L.; Fontcuberta, J. The Direct Magnetoelectric Effect in Ferroelectric-Ferromagnetic Epitaxial Heterostructures. Nanoscale 2013, 5, 8037−8044. (19) Zhang, Y.; Deng, C.; Ma, J.; Lin, Y.; Nan, C.-W. Enhancement in Magnetoelectric Response in CoFe2O4-BaTiO3 Heterostructure. Appl. Phys. Lett. 2008, 92, 062911. (20) Zhang, J. X.; Dai, J. Y.; Chan, H. L. W. Interfacial Engineering and Coupling of Electric and Magnetic Properties in Pb(Zr0.53Ti0.47)O3/ CoFe2O4 Multiferroic Epitaxial Multilayers. J. Appl. Phys. 2010, 107, 104105. (21) Wakabayashi, Y. K.; Nonaka, Y.; Takeda, Y.; Sakamoto, S.; Ikeda, K.; Chi, Z.; Shibata, G.; Tanaka, A.; Saitoh, Y.; Yamagami, H.; Tanaka, M.; Fujimori, A.; Nakane, R. Electronic Structure and Magnetic Properties of Magnetically Dead Layers in Epitaxial CoFe2O4/Al2O3/ Si(111) Films Studied by X-ray Magnetic Circular Dichroism. Phys. Rev. B 2017, 96, 104410. (22) Garcia, V.; Bibes, M. Ferroelectric Tunnel Junctions for Information Storage and Processing. Nat. Commun. 2014, 5, 4289. (23) Xi, Z.; Ruan, J.; Li, C.; Zheng, C.; Wen, Z.; Dai, J.; Li, A.; Wu, D. Giant Tunnelling Electroresistance in Metal/Ferroelectric/Semiconductor Tunnel Junctions by Engineering the Schottky Barrier. Nat. Commun. 2017, 8, 15217. (24) Ohresser, P.; Otero, E.; Choueikani, F.; Chen, K.; Stanescu, S.; Deschamps, F.; Moreno, T.; Polack, F.; Lagarde, B.; Daguerre, J.-P.; Marteau, F.; Scheurer, F.; Joly, L.; Kappler, J.-P.; Muller, B.; Bunau, O.; Sainctavit, P. DEIMOS: A Beamline Dedicated to Dichroism Measurements in the 350−2500 eV Energy Range. Rev. Sci. Instrum. 2014, 85, 013106. (25) Ramos, A. V.; Moussy, J.-B.; Guittet, M.-J.; Gautier-Soyer, M.; Gatel, C.; Bayle-Guillemaud, P.; Warot-Fonrose, B.; Snoeck, E. Influence of a Metallic or Oxide Top Layer in Epitaxial Magnetic Bilayers Containing CoFe2O4(111) Tunnel Barriers. Phys. Rev. B: Condens. Matter Mater. Phys. 2007, 75, 224421. (26) Barbier, A.; Mocuta, C.; Stanescu, D.; Jégou, P.; Jedrecy, N.; Magnan, H. Surface Composition of BaTiO3/SrTiO3(001) Films K
DOI: 10.1021/acsami.8b09499 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Research Article
ACS Applied Materials & Interfaces (44) Tanaka, A.; Jo, T. Resonant 3d, 3p and 3s Photoemission in Transition Metal Oxides Predicted at 2p Threshold. J. Phys. Soc. Jpn. 1994, 63, 2788−2807. (45) Pattrick, R. A. D.; van der Laan, G.; Henderson, C. M. B.; Kuiper, P.; Dudzik, E.; Vaughan, D. J. Cation Site Occupancy in Spinel Ferrites Studied by X-ray Magnetic Circular Dichroism: Developing a Method for Mineralogists. Eur. J. Mineral. 2002, 14, 1095−1102. (46) Matzen, S.; Moussy, J.-B.; Mattana, R.; Petroff, F.; Gatel, C.; Warot-Fonrose, B.; Cezar, J. C.; Barbier, A.; Arrio, M.-A.; Sainctavit, P. Restoration of Bulk Magnetic Properties by Strain Engineering in Epitaxial CoFe2O4 (001) Ultrathin Films. Appl. Phys. Lett. 2011, 99, 052514. (47) Chen, C. T.; Idzerda, Y. U.; Lin, H.-J.; Smith, N. V.; Meigs, G.; Chaban, E.; Ho, G. H.; Pellegrin, E.; Sette, F. Experimental Confirmation of the X-Ray Magnetic Circular Dichroism Sum Rules for Iron and Cobalt. Phys. Rev. Lett. 1995, 75, 152. (48) O’Brien, W. L.; Tonner, B. P. Orbital and Spin Sum Rules in Xray Magnetic Circular Dichroism. Phys. Rev. B: Condens. Matter Mater. Phys. 1994, 50, 12672. (49) Carvallo, C.; Sainctavit, P.; Arrio, M.-A.; Menguy, N.; Wang, Y.; Ona-Nguema, G.; Brice-Profeta, S. Biogenic vs. Abiogenic Magnetite Nanoparticles: A XMCD Study. Am. Mineral. 2008, 93, 880−885. (50) Tileli, V.; Duchamp, M.; Axelsson, A.-K.; Valant, M.; DuninBorkowski, R. E.; Alford, N. M. On Stoichiometry and Intermixing at the Spinel/Perovskite Interface in CoFe2O4/BaTiO3 Thin Films. Nanoscale 2015, 7, 218−224. (51) Oku, M.; Hirokawa, K. X-ray Photoelectron Spectroscopy of Co3O4, Fe3O4, Mn3O4, and Related Compounds. J. Electron Spectrosc. Relat. Phenom. 1976, 8, 475−481.
L
DOI: 10.1021/acsami.8b09499 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX