Charge Transfer Control of Emergent Magnetism at SrMnO3/LaMnO3 Interfaces

Emergent phases at the interfaces in strongly correlated oxide heterostructures display novel properties not akin to those of constituting materials. The interfacial ferromagnetism in (LaMnO3)m/(SrMnO3)n (LMO)m/(SMO)n superlattices (SLs) with antiferromagnetic bulk LMO and SMO layers is believed to be a result of the interfacial charge transfer (CT). By using in situ optical ellipsometry, it is demonstrated directly that CT and emergent magnetism in (LMO)m/(SMO)n SLs are controlled by the LMO/SMO thickness ratio, chosen as m/n = 1 and 2. The enhanced CT in SLs with m/n = 2 favors the high‐TC emergent ferromagnetism with TC = 350–360 K, whereas the reduced CT in m/n = 1 SLs suppresses it yielding TC = 300 K. A complex dependence of the saturation magnetization as a function of interface density Λ = (m + n)−1 with minima at Λ = 0.11 (m/n = 2) and Λ = 0.25 (m/n = 1) was observed and rationalized by the competition of ferromagnetic and antiferromagnetic contributions, originating from the volume of LMO and SMO layers as well as from the LMO/SMO interfaces. The role of epitaxy stress and MnO6 octahedral tilts in the emergent magnetic behavior is discussed.


Introduction
Heterostructures of strongly correlated oxides provide a rich material platform for searching and tuning of emergent interfa-

Sample Preparation and Experimental Details
The [LMO m /SMO n ] K SLs on STO(100) substrates with m/n = 1 and 2, called in the following SL1 and SL2, respectively, with the bilayer thickness, Λ = m + n = 6-20, and repetition number of bilayers, K = 10, were grown using MAD technique as described in details. [15] The growth of SLs has been monitored in situ by optical ellipsometry as presented in Figure 1, where the thickness (D) and time (t) dependences of the ellipsometric phase shift angle, Δ(D = vt), are shown for representative 10/5 and 5/5 SLs. Within the LMO layers one can see a linear increase of the phase shift, Δ, in time/thickness with an almost constant phase shift rate, dΔ/dD = 0.3° u.c. −1 In contrast, within the SMO layers a nonmonotonic behavior was observed with pronounced changes of the dΔ/dD rate within all SMO layers. The insets in Figure 1 with a zoomed view of the phase shift rate behavior demonstrate a sudden increase up to dΔ/dD = 0.5° u.c. −1 occurs within the first 1-2 u.c. of the growing SMO layer. Further on, the rate drops down to small values dΔ/dD = 0.1° u.c. −1 The observed here ellipsometric Δ(D) behavior, being in qualitative agreement with our previous results, [15] was characteristic for all SL1 and SL2 sample series. However, the details of the phase shift rate thickness behavior within the SMO layers close to the interfaces in SL1 and SL2 samples were found to be different and will be analyzed below.
Global structural characterization of SLs was performed by X-ray diffraction (XRD) and X-ray reflection (XRR) using a Bruker D8 diffractometer with Cu-K α radiation. XRD and XRR patterns were simulated by ReMagX program. [18] Microscopic characterization was done using an FEI environmental Transmission Electron Microscope (ETEM) Titan 80-300. TEM Figure 1. Ellipsometric phase shift angle, Δ, as a function of layer thickness, D, measured in situ during the growth of LMO 10 /SMO 5 (top panel, red curve) and LMO 5 /SMO 5 (bottom, blue) superlattices (SLs). The measured Δ(t) diagrams were renormalized to Δ(D = v × t) dependences, using the known constant deposition rate v = 0.3 u.c. s −1 , independently verified by X-ray reflection and transmission electron microscopy (TEM) data. The vertical lines marked the opening and closing of LMO and SMO liquid channels. One can see an almost linear Δ(D) behavior for LMO layers and a nonmonotonous behavior for SMO layers caused by the charge transfer (CT) from the underlying LMO into the growing SMO layer. [15] The insets show the zoomed views of the phase shift rate, dΔ/dD, as a function of layer thickness, which, being proportional to the electronic density in the layer, [15] illustrates the in situ CT across the interface from the LMO into the SMO and an increased electron concentration within the first two u.c. of SMO layers close to the SMO(top)/LMO(bottom) interface (see also Figure S1, Supporting Information). [20] lamellas were prepared in the [100] and [110] directions using a Thermofischer (former FEI) Helios UC focused ion beam instrument with a beam energy of 30 kV. A final cleaning step was performed at low energy (2 kV). A detailed study of crystal structure, including the quantitative analysis of the MnO 6 octahedral tilts in the LMO layers, is presented. [19] Magnetization was measured by using a SQUID magnetometer (MPMS XL, Quantum Design) for temperatures, T = 5-400 K and applied magnetic fields, B = 0-5 T, aligned parallel to the film surface.

Structure and Lattice Strain in the SLs
The high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) images of two representative SL1 (5/5) and SL2 (10/5) samples are shown in Figure S1 (Supporting Information). [20] It demonstrates epitaxially grown LMO and SMO layers with flat and atomically sharp interfaces. Additionally, a high structural quality of layers and interfaces as well as of the whole SL architecture is supported by the measurements and simulations of XRR and XRD (see Figures S2 and S3, Supporting Information [20] ). The thicknesses of the LMO and SMO layers, determined from the XRR, as well as the simulated from XRD values of out-of-plane lattice parameters for individual SMO and LMO layers are presented in Table 1.
One can see a good correspondence between the nominal layer thicknesses (n, m) and those experimentally measured (n exp , m exp ) by TEM and XRR (bilayer thickness) with a deviation between them not exceeding 2-3%. Moreover, the XRD-determined averaged c-axis lattice parameters in SLs, C av , allowed us to estimate the stress states in the LMO and SMO layers and compare the overall stress state of these two series of SLs. One can see from Table 1 that the C av values are almost the same for all samples with the same m/n ratio. However, they differ sizably between the SL1 and SL2 series, having different LMO/ SMO thickness ratio, i.e., C av (SL1) = 0.382 nm and C av (SL2) = 0.3845 nm. Moreover, from the simulations of the XRD patterns ( Figures S2 and S3, Supporting Information) one can evaluate the c-axis lattice parameters of the individual layers, C LMO and C SMO , and calculate the theoretical values of the average C av.th.lattice parameters in SLs as C av.th. = (m × C LMO + n × C SMO )/ (m + n), which were then compared with the measured C avvalues (see Table 1). For the SL1 and SL2 sample series one gets C av.th. (SL1) = (0.3882 + 0.3750)/2 = 0.3816 nm and C av.th. (SL2) = (2 × 0.3892 + 0.3722)/3 = 0.3835 nm. A pretty good correspondence between the measured and theoretically calculated values of the averaged C av -lattice parameters in the SLs reflects the different mass ratio of LMO and SMO in them and confirms that the LMO/SMO thickness ratios are close to the nominal values for all SL samples in the "m/n = 1" and "m/n = 2" series.
Further, the SMO layers in SLs are stronger in-plane tensile stressed as the LMO layers. Namely, the evaluated tensile stress in the SMO layers ε = (C SMO,f -C SMO,bulk )/C SMO,bulk , is equal to ε SMO (SL1) = −1.7% and ε SMO (SL2) = −2.1% (c SMO,bulk = 0.3805 nm [21] ). In contrast, the stress in the LMO layers in SLs, i.e., ε = (c LMO,f -c LMO,bulk )/c LMO,bulk , is estimated to be significantly smaller because the pseudocubic lattice parameter of the LMO layers and also of single LMO films is very close to the in-plane lattice parameter of the STO substrate; the resulting stress values are ε LMO (SL1) = −0.6% and ε LMO (SL2) = −0.3%. Neverthe-less, one can see a difference in the LMO stress states between the SL1 and SL2 series -in the latter LMO layers are much less strained. Interestingly, the overall stress values, given as a sum of the LMO and SMO stresses, i.e., ε SL (SL1) = −1.7 to 0.6 = −2.3% and ε SL (SL2) = −2.1 to 0.3 = −2.4%, are very similar in the SL1 and SL2 samples because the LMO and SMO layers reveal opposite changes.

Magnetism
In Figure 2, we present the temperature and field dependences of the magnetic moment, M(T), measured by SQUID for two representative SL2 (12/6) and SL1 (6/6) samples. One can see a complex behavior with two ferromagnetic transitions with low (T C1 ) and high (T C2 ) Curie temperatures. The T C values were determined from the position of the minimum of the function TCM(T) = (1/M) × (dM/dT) also shown in the inset of Figure 2a. Such M(T) curves with two ferromagnetic transitions were found to be characteristic for all other SL samples in the present study. Moreover, one can see also characteristic differences between the 12/6 (SL2) and 6/6 (SL1) samples, indicating stronger magnetic properties in the 12/6 (SL2) sample. Namely, T C2 = 355 K obtained for the 12/6 SL exceeds considerably the T C2 = 300 K in the 6/6 SL (see Figure 2a) and, moreover, the saturation magnetization in the 12/6 SL exceeds that for the 6/6 SL by more than two times (see Figure 2b). Further, the 6/6 SL shows a significantly larger coercive field H c = 1.8 kOe compared to the H c = 0.5 kOe measured for the 12/6 SL. Finally, both 12/6 and 6/6 SLs display a pronounced exchange bias field, H EB = 90 and 130 Oe, respectively. Concluding, the temperature and field dependent magnetization with two ferromagnetic transitions, shown in Figure 2 for the two chosen SLs, are typical for all other SL2 and SL1 samples in the present study. Moreover, the M(T) and M(H) dependences in SL2 samples reproduce the recently published results on SLs with m/n = 2. [15] In Figure 3, we summarized the results of the temperature and field dependent SQUID magnetization for all present Table 1. The data of X-ray reflectivity (XRR) and X-ray diffraction (XRD) for all superlattices (SLs). Individual thicknesses of layers d SMO and d LMO were obtained from simulations by using of ReMagX program [18] and the mean out-of-plane lattice parameters, C 0 , from the measured XRD patterns. n exp and m exp are evaluated from the XRR and XRD thicknesses of SMO and LMO layers, respectively.
(LMO m /SMO n ) 10 XRR XRD   SL1 and SL2 samples, having different bilayer thicknesses, Λ = m + n, and interface densities, 1/Λ = (m + n) −1 . In addition, we have included the data from ref. [15] to get a complete overview. As one can see from Figure 3a, the T C1 values decrease with decreasing the 1/Λ for both types of SLs and for low interface densities, 1/Λ < 0.1, they approach T C = 160-170 K of single LMO films with a thickness D = 20-40 nm. [15,22] This indicates that the low-T C1 transition originates from a "volume-like" LMO ferromagnetic contribution, which progressively increases with increasing the LMO thickness and decreasing interface density (1/Λ) in SLs. Note that the FM ground state in epitaxial LMO films, in contrast to the AFM ground state in bulk LMO, is known to be stabilized by the epitaxy stress. [23] In contrast, T C2 values, being almost independent on the interface density 1/Λ within both SL series, characterize the onset of the interfacial emergent FM phase. [1] Remarkably, we observed that a high-T C2 emergent ferromagnetism does depend on the m/n ratio as the determined values of T C2 = 355-360 K for the SL2 samples exceed significantly the T C2 = 300 K for the SL1 samples. This suggests that the CT, which was supposed to be the reason of the formation of the emergent FM phase, [15] could be influenced by the LMO/SMO thickness ratio.
In Figure 3b, the saturation magnetization m sat (5K) of SLs is plotted as a function of the interface density 1/Λ; the data of single La 1-x Sr x MnO 3 (LSMO) films as boundary cases for SL1 (x = 1/2) and SL2 (x = 1/3) are added for comparison at 1/Λ = 1. Apparently, one can see a pronounced m sat (1/Λ) dependence with generally reduced values, m sat = 0.5-2.5 μ B Mn −1 , in SLs compared to those for optimally doped bulk LSMO [24] or single LSMO films. [15] Interestingly, a minimum in the m sat (1/Λ) dependence can be seen at 1/Λ = 0.11 (Λ = 9 u.c.) for 6/3 SL2 and at 1/Λ = 0.25 (Λ = 4 u.c.) for the 2/2 SL1. Note that the pronounced minimum for the SL2 samples appears in the vicinity of the transition from an LSMO-like ferromagnetic behavior in SLs with very thin SMO layers (n = 1, 2 u.c.) to an SL-specific magnetism for n ≥ 3, which is manifested by the competition of the interface and bulk-like contributions from the LMO and SMO layers. Indeed, large m sat = 3.3-3.5 μ B Mn −1 comparable with those in the optimally doped bulk LSMO as well as a metallic behavior was observed [15] for SL2 samples with n = 1, 2. At n = 3 a metal-insulator transition occurs in the SL2 system and leads to a strong reduction of m sat probably due to the increase of the antiferromagnetic (AFM) contribution in SLs. This is further supported by a drastic and steep enhancement of coercive field, H c , and appearance of the exchange bias field, H EB , for SL2 samples with n > 3, shown in Figure 3c. For the m/n = 1 SLs a less pronounced minimum in the m sat (1/Λ) dependence seen in the 2/2 SL probably also reflects a transition from a bulk-like behavior similar to the half-doped LSMO for n < 2 to an SL governed magnetism for n > 2. However, at Λ = 4 no MI transition takes place because even the reference half-doped LSMO films are rather insulating and, likely, electronically inhomogeneous. The absence of an MI transition in SL1 samples is additionally illustrated by a monotonous H c (1/Λ) and H EB (1/Λ) behavior shown in Figure 3d.
Thus, the complex m sat (1/Λ) dependences for SL2 and SL1 samples (see Figure 3b) with reduced saturation magnetization and their minima indicate the presence and even the dominance of an AFM phase in SLs with very thin LMO and SMO layers (n = 2-6) and correspondingly large interface density 1/Λ = 0.08-0.25 (u.c.) −1 . Especially pronounced is the AFM phase within the SL1 samples, which displays very low values m sat = 0.4-1.2 μ B Mn −1 . The AFM contributions originate most probably from the SMO layers, which start to acquire their bulk magnetic properties for thicker layers n > 3. [15] Another source of the AFM contribution could be provided by the LMO(top)/ SMO(bottom interface) as according to polarized neutron scattering [15] the first 3 u.c. of the LMO grown on top of the SMO also possess an AFM ground state with a nearly zero magnetization. This means that SLs with a thicker SMO (n ≥ 3 u.c.) and LMO (2n > 6 u.c.) layers and relatively high interface density 1/Λ = 0.1-0.2 (u.c.) −1 display lower values of m sat , dominated mostly by the interfacial FM contribution, as the AFM ordered LMO and SMO layers likely do not contribute to m sat . By further increasing the layer thickness and further reducing the interface density the LMO finally becomes an FM and m sat of an SL increases. For an SL with thick LMO and SMO layers and vanishingly small interface density (1/Λ→0), which as well results in a vanishingly small interfacial FM contribution, the m sat should approach the volume-like LMO value as bulk SMO being in the AFM ground state do not contribute to m sat . Nevertheless, the coercive field and exchange bias fields remain large (see Figure 3c

Charge Transfer
The main result of the present study is that the emergent magnetism in LMO m /SMO n SLs can be influenced by the "m/n" thickness ratio of the constituting layers: the large "m/n = 2" ratio favors ferromagnetism and results in a high-T C emergent phase while the small "m/n = 1" ratio suppresses it. Taking into account that the reason of the emergent magnetism is a CT across the SMO/LMO interface [15] and considering previous results of Garcia-Barriokanal et al., [7] one can suggest that CT is enhanced in the SL2 and is suppressed in the SL1 samples. To examine this issue we have analyzed the ellipsometry time (t)/ thickness (D) diagrams measured in situ during the growth of SL2 and SL1 samples and shown exemplified for 5/5 and 10/5 SLs in Figure 1. Our analysis is based on the phase shift rate, dΔ/dD, which reflects the electron density and its change in the growing film. [15] Indeed, the thin LSMO films [15] and LMO layers in SLs (see Figure 1) with well-defined electronic properties show a linear increase of delta during growth. This can be described by a phase shift rate dΔ/dD, which stays constant during the layer growth and provides information on the electronic properties of the growing layer as will be discussed in the following. Note, that ellipsometry as an optical/infrared reflection technique is generally sensitive to electronic properties of a material probed at a special excitation photon energy, [25] here E = 1.96 eV (light wavelength λ = 632.8 nm). To elaborate the relationship between the phase shift rate and charge density we modeled the in situ ellipsometry signal of a growing multilayer using the Fresnel equations [25] and have shown that it is determined by the complex refractive index of the top layer (see Supporting Information [20] ), which in turn is related to its electronic properties. With refractive indices of single LSMO films with Sr doping 0 < x < 1, evaluated from measurements of in situ ellipsometry (see Figure S4, Supporting Information [20] ), the values of phase shift rate dΔ/dD were determined from the Fresnel equations for very thin LSMO layers with D ≤ 12 u.c., which are relevant for the SLs (see Figure S5, Supporting Information [20 ). One can see a characteristic dependence the phase shift rate on "x" with a pronounced maximum of dΔ/dD for optimally doped LSMO, which explains the enhanced phase shift rate at the SMO(top)/ LMO(bottom) interfaces due to the CT. [15] The simulations also reproduce the relatively large values of dΔ/dD = 0.3° u.c. −1 , observed in electron-rich LMO (see Figure 1) and small dΔ/dD = 0.1° u.c. −1 seen in the electron-poor SMO layers. It is important to note that the photon energy E = 1.96 eV, used in our ellipsometry setup, fits almost exactly the excitation energy (E = 2 eV) of the intersite Mn 3+ →Mn 3+ optical transitions. [26] Considering the fact that spectral weight of these transitions is very sensitive to Sr-or Ca-doping [27] or in other words to the Mn 3+ /Mn 4+ ratio, the sensitivity of the ellipsometric phase shift rate to the electron/ hole density in manganites can be rationalized.
Being aware that we cannot directly determine the amount of charge transferred across the interface, we carried out a comparative analysis of the phase shift rate within the SMO layers in all present SL samples with m/n = 1, 2 with the main aim to quantify the differences in the CT length λ TF within these two series of SLs. In Figure 4, we present a comparison of the phase shift rate within the SMO layers in SL2 and SL1 samples, having the same thickness of SMO, e.g. we compare the 10/5 and 5/5 SLs. One can see that all studied SL2 samples (red symbols) display a sizably broader distribution of the phase shift rate or, in other words, of charge spreading into the SMO layers than that measured in the SL1 (blue symbols). For example, the CT length, λ TF , estimated in the 6/6 and 12/6 SLs was λ TF = 1 and 2 u.c., respectively. For other SLs with thinner SMO layers (n = 3, 4, 5 u.c.) the progressively smaller λ TF values were obtained. Interestingly, the relative difference in δλ TF between the SL1 and SL2 samples normalized to the thickness of SMO layers n = 3, 4, 5, 6, i.e., δλ TF /n, or charge spreading within SMO was found to be nearly the same for all SL2 samples, i.e., ≈0.5 u.c./3 = 0.166; ≈0.65 u.c./4 = 0.165; ≈0.8 u.c./5 = 0.16, and 1 u.c./6 = 0.166. These data indicate, likely, a very similar "effective electron doping" for all SL2 samples, having different SMO thickness, "n," but the same ratio m/n = 2. This supports our suggestion that the emerging high- T C2 ferromagnetic phase is caused by the enhanced CT within the SMO layers in m/n = 2 SLs.
An additional way to detect the CT across the interface can be the estimation of the oxidation states of the Mn ions inside the LMO and SMO layers in SLs. Indeed, the electrostatic mismatch at the SMO/LMO interface with the formally charged (La-O) +1 /(Mn-O 2 ) −1 layer and formally neutral (Sr-O) 0 /(Mn-O 2 ) 0 layer, could be relaxed by the CT. Assuming that LMO loses electrons and the SMO acquires them, this should lead to the modified oxidation states of Mn 3+δ within the LMO and Mn 4-δ (SMO). Moreover, also the layers above and below the interface could be involved because of the charge redistribution within the bilayer. The most accurate tool for estimation of Mn oxidation states was shown [28,29] to be electron energy loss spectroscopy (EELS) at the O-K edge and determination of the energy distance between the O-K b and O-K a peaks, ΔE ba = E Oka − E OKb . This procedure, compared to the L 3 /L 2 ratio [30,31] measured at the Mn-L 2,3 edge, is less sensitive to the noise. The high resolution EELS behavior for the 6/6 and 12/6 SLs, shown in Figure 5a,b, respectively, demonstrates how the spectra close to the O-K and Mn-L 2,3 edges are distributed along the growth direction of SLs. One can see qualitative differences between the LMO and SMO layers in SLs in agreement with our previous results. [15] The SMO layers are manifested by an energy shift of the O-K edge (mostly of the central peak) and by the intensity changes of the Mn-L 2,3 edge due to the changes in the L 3 /L 2 ratio. The obtained EELS spectral behavior is generally similar to that reported for bulk manganites and thin films previously. [28][29][30][31] In Figure 5c,d, we present the EELS spectra at the O-K edge, measured in the middle of the LMO (Figure 5c) and SMO (Figure 5d) layers for 12/6 and 6/6 SLs as well as for the reference LMO/STO and SMO/STO single films. The measured EELS spectra contain typical features, called O-K a and O-K b peaks, [28,29] the energy distance between which ΔE ba has been demonstrated to be very sensitive to the oxidation states of Mn ions [28] as well as for stress state of the films as demonstrated for the SMO films grown on different substrates. [29] The peaks were fitted by Gauss curves as shown in Figure 5c,d and the values ΔE ba were determined. As one can see in Figure 5e, by comparing  [28] obtained for La 1-x Ca x MnO 3 manganites. the data on our samples with the data on La 1-x Ca x MnO 3 measured by Varela et al., [28] the reference LMO/STO and SMO/STO films do possess the nominal oxidation states, i.e., Mn 3+ and Mn, [4] respectively. The values of ΔE ba obtained in the middle of the LMO and SMO layers in SLs, i.e., "far away" from interfaces (see Figure 5c,d, respectively) indicate modified oxidation states. Namely, oxidation states of Mn in LMO layers inside the SLs are relatively close to Mn 3+ , i.e., V LMO (12/6) = + 3.05 and V LMO (6/6) = +3.1. In contrast, those for the SMO layers deviate significantly from Mn 4+ : V SMO (12/6) = +3.5 and V SMO (6/6) = +3.7. Note, the obtained here small deviations from the nominal Mn 3+ in LMO layers, being close to the uncertainty of ±0.15 given by the L 3 /L 2 method, [30] cannot be reliably determined by it. However, the determined Mn oxidation states in SMO layers exceed this uncertainty. In principle, the oxidation state of Mn in SMO films can be also influenced by strain-induced oxygen deficiency as demonstrated by Agrawal et al. [29] Comparing the ΔE ba values for SMO layers in our SLs with their results, similar oxidation states should be resulted from the 2.6% and 3.7% strain in the SMO layers for the 6/6 and 12/6 SLs, respectively. However, according to the XRD the stress values in SMO layers are significantly smaller, i.e., 1.7% (6/6) and 2.1% (12/6) and, moreover, the oxygen deficiency was not detected in our SMO layers. Thus, we conclude that the stress alone cannot lead to the observed strong deviations of Mn oxidation states in SMO. The most probable reason seems to be the in situ LMO-to-SMO CT, followed by a redistribution of the electron density within the underlying LMO layers.

Discussion
We performed a systematic study of the magnetism in LMO m / SMO n SLs (m/n = 1,2; m, n = 2-10 u.c.) and demonstrated its correlation with electronic/structural reconstructions at the interfaces analyzed by in situ ellipsometry and ex situ EELS. The observed high-T C emergent magnetic behavior in m/n = 2 SLs was assigned to an enhanced CT within ≈2 u.c. at the SMO/ LMO interfaces. The latter were optically/electronically modeled by assuming the formation of an optimally doped LSMO (x = 1/3) phase at the interface. Emergent behavior observed here and elsewhere [15] in the MAD grown LMO 2n /SMO n SLs is quite different from that found in the MBE grown SLs, [10,32] which possess strongly asymmetric interfaces and do not show high-T C emergent phase. Here, we would like to note that interfaces in MAD-grown SLs are flat and quite similar-the SMO(top)/ LMO(bottom) is atomically sharp with an XRR roughness, σ SMO/LMO = 1 u.c., and the LMO(top)/SMO(bottom) looks less sharp (σ LMO/SMO = 1.5 u.c.). According to the ellipsometry and polarized neutron scattering [15] the emergent high-T C FM phase is located at the sharp SMO/LMO interface, which can be modeled by electronic properties of optimally doped LSMO. This indicates that the electronic reconstruction is responsible for the emergent high-T C FM phase. A chemically less sharp and likely intermixed LMO/SMO interface with accompanying chemical disorder is not relevant for high-T C magnetism. Moreover, assuming only a chemical intermixing at the interfaces, a half-doped interfacial LSMO (x = 0.5) phase should be present in both m/n = 1 and 2 SLs. This, however, contradicts (see Figure 3a) the experimentally observed systematic difference in the emergent ferromagnetism in these SLs.
The estimated from the ellipsometry CT lengths into the SMO in SLs, e.g. λ TF (6/6) = 1 u.c. and λ TF (12/6) = 2 u.c., can be related to the EELS-determined average Mn oxidation states V LMO and V SMO assuming a redistribution of the electron density within the layers as follows: V LMO = [3(m − λ TF ) + 4λ TF ]/m = 3 + λ TF /m and V SMO = [4(n − λ TF ) + 3λ TF ]/n = 4 − λ TF /n. For the LMO in 12/6 and 6/6 SLs one gets a similar Mn oxidation sates, V LMO (12/6) ∼ V LMO (6/6) = +3.17, because of a factor of two larger CT length for the 12/6 SL than that for the 6/6 SL. These values are in line with those measured by EELS (see Figure 5e) in the middle of the LMO layers in SLs, =+3.1(12/6). Moreover, the average Mn oxidation states in SMO are: V SMO (12/6) = 3.67 and V SMO (6/6) = 3.83, which also agree quite good with the EELS-determined values for SMO, i.e., +3.5 (12/6) and +3.7 (6/6). Thus, the control of CT by the "m/n" thickness ratio, its relevance for the emerging magnetism and its enhancement in the m/n = 2 SLs is qualitatively supported both by in situ ellipsometry and ex situ EELS measurements.
Another important issue is whether the CT and stress state in the studied SLs are related to each other. For example, as was shown by Garcia-Barriocanal et al. [7] the CT leads to almost fully relaxed LMO layers in the LMO/STO SLs. As we have shown above in Section 3 the m/n = 1 and 2 SLs possess a very similar overall tensile stress state with SMO layers much more strained than those of LMO. It is interesting to note that an enhanced CT in the SLs with m/n = 2 and decrease of stress in the LMO layers likely correlate with each other. At this point we speculate that the transferred electrons into the SMO help to relax the LMO by reducing the tetragonal distortion by means of MnO 6 octahedral tilt mechanism. According to the recent study of Meyer et al. [19] the main structural difference between the present SL1 and SL2 samples is the existence of octahedral tilts in the LMO layers within the m/n = 2 SLs and their absence in the m/n = 1 SLs. The presence of tilts is consistent with an increased averaged C av -lattice parameter measured for the SL2 samples (see Table 1) and, hence, with a decreased tetragonal distortion in them. Note, that Kan et al. [33] also reported an example of anisotropic strain relief in the SrRuO 3 films by engineering of octahedral rotations at the interface between SrRuO 3 and GdScO 3 substrate. However, the microscopic scenario of the strain relief by octahedral rotations and the role of charge and orbital reconstructions in it still needs a detailed theoretic description.
We believe the m/n ratio could affect the magnitude of the CT due to following mechanisms: a) mechanic strain or "mass" effect-the larger the mass ratio the larger the influence of LMO on SMO; b) structural effect actuated by different types of interfaces, i.e., Mn−O−Mn tilt-mismatched in the m/n = 2 SLs and mismatch-free the m/n = 1 SLs; and c) electrostatic mismatch effect quantified by the difference in capacities of charge reservoirs (electron-rich LMO and electron-poor SMO) might directly drive the interfacial CT. In addition, we would like to note that such a direct "electrostatic mismatch" effect on CT, being proceeded already during the growth at high temperatures, T sub = 900-950 °C, plays a dominant role in our opinion. The role of other two factors (a) and (b) and their casual relationship in the realization of CT seems to be not clear up to now.