Atomic-Scale Quantitative Analysis of Lattice Distortions at Interfaces of Two-Dimensionally Sr-Doped La2CuO4 Superlattices.

Using spherical aberration corrected high-resolution and analytical scanning transmission electron microscopy, we have quantitatively studied the lattice distortion and the redistribution of charges in two-dimensionally strontium (Sr)-doped La2CuO4 superlattices, in which single LaO planes are periodically replaced by SrO planes. As shown previously, such structures show Tc up to 35 K as a consequence of local charge accumulation on both sides of the nominal SrO planes position. This is caused by two distinct mechanisms of doping: heterogeneous doping at the downward side of the interface (space-charge effect) and "classical" homogeneous doping at the upward side. The comparative chemical and atomic-structural analyses reveal an interrelation between local CuO6 octahedron distortions, hole spatial distribution, and chemical composition. In particular we observe an anomalous expansion of the apical oxygen-oxygen distance in the heterogeneously doped (space-charge) region, and a substantial shrinkage of the apical oxygen-oxygen distance in the homogeneously doped region. Such findings are interpreted in terms of different Jahn-Teller effects occurring at the two interface sides (downward and upward).

Introduction

The possibility of tuning oxide functionalities by the insertion of an interface or active layers has been frequently investigated in recent years.[1−3] Tuning of properties via local interface effects at grain boundaries, at epitaxial contact between different phases,[4−7] and via delta-doping[8,9] has been achieved. This has resulted in the improvement of electrical performance in ionic and mixed conductors[10] and in the emergence of novel effects, ranging from high-temperature superconductivity[10] to electrical conductivity[11] and magnetism.[12]

A major challenge in the study of oxide interface effects arises from the complex interplay between several local effects. Particularly, the rearrangement of ionic and electronic charges[3,13,14] and the distortion of the crystal structure at the interface[15−17] play a crucial role.

Using atomic layer-by-layer oxide molecular beam epitaxy (ALL-oxide MBE), we fabricated two-dimensionally strontium (Sr)-doped La2CuO4 (LCO) superlattices on LaSrAlO4 (LSAO) substrates. Active (negatively charged) interfaces were intentionally inserted by replacing SrO dopant planes for full atomic planes of LaO with a predefined periodicity.[18] By appropriately choosing the spacing between the dopant planes, the resulting electrical properties of such heterostructures exhibit high-temperature superconductivity (HTSC) up to ∼35 K. It has been shown that HTSC in these heterostructures is a consequence of the local charge accumulation occurring on both sides of the doped planes as a consequence of different mechanisms of doping: (i) heterogeneous doping at the downward side and (ii) “classical” homogeneous doping at the upward side of the interface. Here “heterogeneous doping” means that hole accumulation occurs to compensate for the spatially confined ionic negative charge stemming from the SrO layer (i.e., a space–charge region is formed as a consequence of two-dimensional doping). Most importantly, in this case electronic and dopant (Sr) spatial distributions are decoupled. “Homogeneous doping” refers to the local compensation of Sr (zero-dimensional) point defects by electron holes.[18] This situation, which is triggered by the highly asymmetric Sr distribution resulting from the growth kinetics, is therefore characterized by the presence of two spatially separated doping modes. It offers us a unique opportunity to investigate the interrelation between local functional and structural properties including layer-dependent superconductivity, local lattice and CuO6 octahedron distortion, chemical stoichiometry, and charge distribution in one sample.

Starting from our previous work, the achieved goal of this study was to use a combination of ALL-oxide MBE synthesis and scanning transmission electron microscopy (STEM), including atomic-resolved imaging, electron energy-loss spectroscopy (EELS), and energy-dispersive X-ray spectroscopy (EDXS), for a detailed investigation of the local chemistry and microstructure of two-dimensionally Sr-doped LCO interfaces. This enabled us to visualize the interrelation between the mobile charge carrier distribution and the local lattice distortions occurring in proximity to the SrO planes.

Experimental Section

Two-dimensionally Sr-doped LaCuO4 superlattices were grown on LaSrAlO4 (001) substrates (Crystec GmbH) using atomic layer-by-layer oxide molecular beam epitaxy (ALL-oxide MBE, DCA Instruments).[19] The superlattices were synthesized at 600 °C and an ozone pressure of 3 × 10–5 Torr. During growth, the samples were monitored in situ using reflection high-energy-electron diffraction (RHEED, see Supporting Information Figure S1). More details of the growth conditions and basic characterization can be found in an earlier study.[18]

Cross-sectional TEM specimens were prepared by a standard procedure which included mechanical grinding, tripod polishing, and argon ion beam milling in a stage cooled with liquid nitrogen. Before STEM experiments, samples were plasma-cleaned in a Fischione plasma cleaner in a 75% argon–25% oxygen mixture for 4 min to eliminate hydrocarbon surface contamination. STEM investigations were performed using a JEOL JEM-ARM 200CF scanning transmission electron microscope equipped with a cold field emission electron source, a DCOR probe corrector (CEOS GmbH), a 100 mm2 JEOL Centurio EDX detector, and a Gatan GIF Quantum ERS electron energy-loss spectrometer. The microscope was operated at 200 kV and a semiconvergence angle of 20.4 mrad, resulting in a probe size of 0.8 Å (1 Å for the EELS and EDXS analyses). Collection angles of 75–309 mrad and 11–24 mrad were used to obtain the HAADF and ABF images, respectively. A collection semiangle of 68.5 mrad was used for EELS measurements. To improve the signal-to-noise ratio (SNR) and minimize the image distortion of high-angle annular dark-field (HAADF) and annular bright field (ABF) images, 10 serial frames were acquired with a short dwell time (2 μs/pixel). Multiple images were then aligned and superimposed. To reduce the noise in the data, the STEM images and EELS maps were processed using the multivariate weighted principal component analysis (PCA) routine (MSA Plugin in Digital Micrograph) developed by M. Watanabe.[20] The STEM images were simulated using a multislice method implemented in QSTEM image simulation software.[21] The optical parameters used for the simulation were the same as the experimentally obtained values. The thickness of the LCO supercell used in the simulation was 15 nm.

Results and Discussion

Figure illustrates the microstructure of a two-dimensionally Sr-doped LCO heterostructure, the composition of which can be formally described as (1 × SrO-LaO-CuO2 + 7 × LaO-LaO-CuO2) × 8 on the LSAO (001) substrate. Figure a and Figure b,c (magnifications of the areas highlighted in red and blue in Figure a, respectively) are HAADF STEM images of this sample. The observation direction is parallel to the crystallographic [100] direction of LSAO. The LSAO substrate has a tetragonal structure with lattice parameters a = 3.756 Å and c = 12.635 Å. LCO exhibits an orthorhombic structure (space group Cmca) with lattice parameters a = 5.357 Å, b = 5.400 Å, and c = 13.156 Å at room temperature.[22,23] Neglecting the small orthorhombic distortion, LCO can be regarded as pseudotetragonal with lattice parameters of a = 3.803 Å and c = 13.156 Å. As can be seen from the atomic resolution HAADF image in Figure b, high-quality superlattices show epitaxial growth on LSAO with orientation relationships: (001)LCO//(001)LSAO, [100] LCO//[100]LSAO, and [010] LCO//[010]LSAO (in the pseudotetragonal system). Because the HAADF image intensity is approximately proportional to Z1.7–2 (with Z being the atomic number)[24,25] the intensity drop corresponds to the Sr-doped regions (ZLa = 57 and ZSr = 38). The superlattices exhibit a very high structural quality and have the nominal superlattice periodicity (also see the RHEED in situ analysis performed during superlattice growth in Figure S1). In particular, the Sr-doped areas are coherent and free from extended defects (Figure c). Most interestingly, the average image intensity plotted along the c-axis crystallographic direction (Figure d and 1e) shows that, in the Sr-doped region, the image intensity exhibits a relatively sharp intensity drop followed by a gradual increase. This finding points to an asymmetric Sr redistribution. The Sr diffuses into the LCO matrix in the growth direction, leading to a highly asymmetric dopant profile.[18]

Figure 1

Atomic structure of the superlattices. (a) Low-magnification HAADF-STEM image of (1 × SrO-LaO-CuO2 + 7 × LaO-LaO-CuO2) × 8 superlattices grown on LSAO (001) substrate. The image was taken along the crystallographic [100] direction of LSAO. Sr-doped areas exhibit a darker contrast, while undoped LCO layers appear brighter. Atomic resolution HAADF images of (b) the LCO/LSAO interface and (c) of Sr-doped areas. (d and e) HAADF image intensity profiles. The green region indicates the LSAO substrate and the blue region highlights Sr-doped areas. The red line in part (e) is a guide to the eye, illustrating the asymmetric Sr concentration profile.

A detailed study of the Sr distribution and concentration at the doped interfaces was performed by EDXS and EELS analyses. In Figure , the Sr-L EDXS and Sr-L2,3 EELS line-scan profiles show that Sr is redistributed over a few layers in LCO and has an asymmetric concentration profile with an extended tail along the growth direction. In Figure a we show, as a reference, the HAADF image from the area analyzed by EELS. The atomically resolved Cu, La, and Sr elemental maps, obtained from EELS spectrum imaging across a Sr-doped interface, are displayed in Figure b–e. From the laterally averaged La, Cu, and Sr signals, the asymmetric Sr distribution is clearly resolved, with Sr replacing La up to about 2 LCO unit cells away from the nominally doped atomic plane. Thus, the estimated Sr redistribution lengths (L) are Lbottom = 0.9 ± 0.2 nm and Ltop = 2.3 ± 0.4 nm for the downward and the upward sides of the interface, respectively (see also Supporting Information Figure S2). The same conclusion can be drawn from Figure f, in which the averaged line profiles for the different elements are shown. It is worth noticing that Yacoby et al.[26] found a broken inversion symmetry in La2–SrCuO4 films, i.e., Sr concentration was found to be systematically lower in the (La,Sr)O atomic layer just below the CuO2 layer and much higher in the (La,Sr)O layer above it. In our Sr-doped regions, no such broken inversion symmetry was observed; rather the Sr concentration drops gradually at each LaO layer along the growth direction, as shown in Figure g (the blue signal shows Sr concentration as a function of distance, the red shows the multi-Gaussian peak fitting at each LaO atomic plane).

Figure 2

Atomic resolution EELS spectrum image of the Sr-doped region. (a) HAADF image of the interface simultaneously recorded with EELS mapping. (b– d) Atomic column resolved Cu-L edge, La-M edge, and Sr-L edge elemental maps obtained by fitting the EELS data to the reference spectra using a multiple linear least-squares fitting procedure. (e) A red-green-blue map of Cu (red), La (green), and Sr (blue). (f) Averaged line profiles from the EELS spectrum imaging. The asymmetric Sr profile is clearly visible. (g) The multi-Gaussian peak fitting (in red) shows that Sr concentration (in blue) gradually decreases along the growth direction at each LaO atomic plane.

Next, we turned our attention to the hole distribution across the doped interfaces. As reported previously, the pre-edge feature of the O−K edge is very sensitive to the hole concentration,[27,28] enabling the local determination of the hole concentration in the superconducting phase.[29] The O–K near-edge fine structure was investigated by EELS. As shown in Figure a typical O–K edge spectra recorded in the Sr-doped region (red) and in the LCO region (black) can be readily distinguished: a pre-edge feature at around 528 eV (in yellow), which is attributed to transitions from the O 1s core level to hole states with p symmetry in the valence band,[30,31] is clearly seen in the former. The black curve shows no detectable prepeak. The intensity of the pre-edge peak has been quantified by multi-Gaussian peak fitting using a nonlinear least-squares (NLLS) routine for all spectra in the line-scan profile across several interfaces. The Gaussian peaks used for the NLLS fitting procedure are shown in Figure a.

Figure 3

Concentration of holes and Sr2+ in the Sr-doped region. (a) EELS oxygen-K edge spectra from a Sr-doped LCO region (red) and from undoped LCO (black). The O–K pre-edge intensity (yellow area) is present in the former. The Gaussian peaks used for NLLS fitting are shown. (b) Overlay of electron hole and Sr concentration profiles as a function of the distance from the nominal Sr-doped layer position. The holes were quantified by multi-Gaussian peak fitting of the O–K edge in the energy-loss range 525–540 eV. In the top x-axis, P refers to the distance from the nominal position of the doped layer, expressed in number of CuO2 planes (plus and minus signs refer to the upward and downward side of the interface, respectively). The right panel of part b shows the generic phase diagram of HTS, i.e., the dependence of Tc on the hole concentration by the empirical formula Tc = Tcmax[1 – 82.6(p – 0.16)2], where p is the hole concentration.[35,36] From this, one can infer the corresponding Tc of any specific CuO2 plane.

To quantify hole and Sr concentrations per building block of La2–SrO4 (consisting of one CuO2 layer and two surrounding La(Sr)O layers), we averaged the EELS intensity profiles (O-K edge prepeak and Sr-L2,3 peaks) for line scans over different Sr-doped regions. Subsequently, the amplitude of the hole profile was scaled to satisfy the charge neutrality condition (i.e., the concentration of electron holes integrated over the interface region was set equal to the total Sr content). By doing this, we obtained the Sr and hole concentrations as functions of the distance from the nominal SrO plane position shown in Figure b (red and blue curves for holes and Sr, respectively). On the top x-axis each CuO2 plane is numbered based on its position with respect to the nominally doped plane. In the two profiles one can again observe the pronounced asymmetry of the Sr concentration whereas, most interestingly, the hole profile is symmetric around the nominal position (x = 0) of the SrO layer. Such a finding indicates that the distribution of the holes is remarkably different from the distribution of the Sr dopant atoms. In particular, on the downward side of the interface, the hole concentration decreases much more gradually than the Sr concentration. This highlights that the region with CuO2 atomic plane numbers P = −4, –3, and −2 is doped via a “nonconventional” mode, i.e., by heterogeneous (two-dimensional)[18] doping. The highly confined Sr dopant layer acts as a negatively charged region, which is electrically compensated via the formation of a hole accumulation layer (space–charge effect) on the downward side of the interface. On the upward side of the interface, the formation of a space–charge region is hindered by the smeared Sr profile, i.e., the Sr distribution width is larger than the screening length. In this case, the hole concentration follows the Sr2+ ion concentration as in conventional homogeneous (one-dimensional)[1,18] doping. Complementary investigations (e.g., zinc-tomography, conductivity experiments), which confirmed such a model, are reported elsewhere.[18] Notably, the decoupling between the Sr2+ dopant and the electron holes, resulting from the difference in the chemical potentials of the metallic and insulator layers, has been observed in a related system, namely at the La1.55Sr0.45CuO4(metallic) + La2CuO4(insulator) bilayer interface.[32−34]

To evaluate the local atomic distances across the Sr-doped interfaces, we used simultaneously acquired high-resolution HAADF and ABF images, which enabled us to image all the elements (La/Sr, Cu, and O) constituting the crystal structure.[37,38]Figure a presents the atomically resolved overlay of HAADF (blue) and ABF (red) images of an area covering four unit cells around the doped plane. Taking advantage of the simultaneous acquisition, we were able to measure the relative atomic positions of the cations and anions. Most importantly, the oxygen atomic columns are clearly resolved as dark dots on red background. The position of the nominal Sr-doped plane (indicated by the yellow arrows in Figure a and Supporting Information Figure S4) was obtained from the HAADF intensity profile which is sensitive to the A-site cations. To quantitatively analyze the local lattice distortion, we mapped all the atomic positions from the HAADF and ABF images by first locating the center-of-mass and then iteratively processing 2D Gaussian fitting refinement for each atomic column (the results are shown in Figure S4). After having located and identified the atomic columns, we measured the La–La and O–O distances from the HAADF and ABF images, respectively. The reported interatomic distance are averages over 13 unit cells of the pseudotetragonal perovskite lattice along the basal direction (in-plane direction). Figure shows the variations of the La–La spacing (b) and the O–O spacing (c) for each LCO perovskite block, as a function of the distance from the nominal position of the SrO layer, which is marked by the dotted line (x = 0). The integer values on the top of the plot correspond to the CuO2 plane belonging to the LCO blocks under consideration. All the distances have been measured after calibration using the interatomic distances of the substrate structure (see Supporting Information Figure S3). In Figure b, the in-plane (d1) and out-of-plane (d2) La–La atomic distances are shown. The values of d1 are comparable with the in-plane lattice parameter of the substrate, suggesting that the film is under epitaxial compressive strain (see also Figure S3). This is confirmed by the fact that no extended defects, e.g., misfit dislocations, which typically result from epitaxial strain release, were observed (Figure a). The d2 values are lower than the d1 values and exhibit a maximum in correspondence with the highest Sr content (at P = −1), indicating that the Sr doping slightly expands the lattice of La2CuO4 along the c-axis. This finding is in good agreement with the literature data. For example Radaelli et al.[39] report for bulk LSCO an expansion of about 2% (from d2 ≈ 3.63 Å to d2 ≈ 3.70 Å) upon LSCO composition change from x = 0 to x = 0.35. In our case, d2 varies from 3.53 Å (at P = −4, where we expect the lowest Sr content) to 3.58 Å (at P = −1).

Figure 4

High resolution STEM image and quantitative analyses of structure distortion in the Sr-doped region. (a) Overlay of simultaneously acquired HAADF (blue) and ABF (red) images of one periodic structure of the Sr-doped region showing the cationic and anionic positions. The inset shows the simulated STEM image (marked with a yellow rectangle). The yellow arrows on the image indicate the nominal position of the SrO layer. (b) La–La atomic column spacing along the in-plane (d1) and out-of-plane (d2) directions as a function of distance from the nominal Sr-doped layer. (c) O–O spacing along in-plane (basal, dB) and out-of-plane (apical, dA) directions as a function of distance from the nominal Sr-doped layer. The error bars give the 95% confidence interval (corresponding to 2 times the standard error) of the average of 13 unit cells of the pseudotetragonal perovskite lattice along the basal direction.

In Figure c, we show the variations of the O–O distances (basal or in-plane (dB, black line) and apical or out-of-plane (dA, red line)) for each LCO perovskite block. The basal distance values correspond to those of the substrate as a consequence of epitaxial strain, while the apical distances are systematically larger, meaning that the CuO6 octahedra are elongated along the c-axis. This can be explained in terms of Jahn–Teller (JT) effect.[40,41] In contrast to dB, which stays constant through the interface, dA varies significantly near the Sr-doped region. The dA distance exhibits a maximum value at the P = −2 CuO2 atomic plane (dA ≈ 4.86 Å) and a minimum value for P = 1 (dA ≈ 4.57 Å). The variation from the values measured two to three building blocks away from the Sr-doped plane (4.72–4.78 Å) is substantial whereas, far from the interface, changes in dA are as small as the measurement accuracy (about 4 pm).[42,43] The same quantitative trend was obtained from similar analyses carried out at different locations along the Sr-doped layers (another example is presented in Supporting Information Figure S5). Interestingly, such a variation cannot be attributed to structural modifications stemming from the Sr distribution: if one takes as a reference the reported values of the distance between Cu and apical-O, i.e., half of the O–O distance under consideration (assuming that the Cu atoms are not significantly displaced from the center of the Cu–O octahedron), one expects small variations (≈0.05 Å) upon Sr content change and, most importantly, a monotonic shrinkage of dA upon Sr increase.[17,39,41] While this argument can be used to explain the measured minimum at P = 1, both the extent of the variation (about 0.2 Å from P = −2 to P = 1) and the presence of a maximum at P = −2 (where the Sr concentration is negligible) clearly indicate the occurrence of a structural anomaly on the downward side of the interface where the heterogeneous doping mode is active. Remarkably, this anomaly occurs at the same position (P = −2) at which, according to complementary investigations, the optimal doping level for superconductivity is reached.[18] It is worth noting here that in a related system, metallic/insulator bilayers, in which the occurrence of interfacial high-temperature superconductivity was attributed to electron transfer, an “anomalous expansion” of the Cu–O distance at the interface was also found.[17] Such an asymmetric apical-oxygen displacement suggests a different JT effect at the two sides of the Sr-doped planes: (i) an anti-JT effect at the upward interface where holes are located in both x2–y2 and z2 orbitals and (ii) an enhanced JT effect at the downward interface, where holes are located mainly in x2–y2 orbitals.[44]

Conclusion

We demonstrated the feasibility of using spherical aberration-corrected STEM to quantitatively describe the charge distribution and the local atomic distances in two-dimensionally Sr-doped LCO superlattices, in which high-temperature superconductivity arises at the interface as a consequence of charge redistribution. By HAADF imaging and EELS spectroscopy, we found the cationic dopant profile to be highly asymmetric: abrupt at the downward side of the interface and smeared in the growth direction. Conversely, the hole distribution, as measured by EELS spectroscopy, is symmetric across the interface and is decoupled from the dopant profile at the downward interface. This indicates that hole doping is achieved on the two sides of the Sr-doped plane by two distinct mechanisms, heterogeneous doping and homogeneous doping at the downward and the upward sides of the interface, respectively. The interatomic structure analysis allowed us to precisely define both the cationic (La) and anionic (O) sublattice positions, highlighting that, whereas the former exhibits only a small dependence on the charge distribution, the latter is significantly affected. In particular, the “soft” Cu-apical O bond, which is known to strongly affect Tc, shows a dramatic increase (as large as 15 pm) from the metallic nonsuperconducting CuO2 plane (P = 1) to the optimally doped CuO2 plane (P = −2) in the space–charge region.