Ruddlesden–Popper Faults in NdNiO₃ Thin Films

Abstract

The NdNiO₃ (NNO) system has attracted a considerable amount of attention owing to the discovery of superconductivity in Nd_0.8Sr_0.2NiO₂. In rare-earth nickelates, Ruddlesden–Popper (RP) faults play a significant role in functional properties, motivating our exploration of its microstructural characteristics and the electronic structure. Here, we employed aberration-corrected scanning transmission electron microscopy and spectroscopy to study a NdNiO₃ film grown by layer-by-layer molecular beam epitaxy (MBE). We found RP faults with multiple configurations in high-angle annular dark-field images. Elemental intermixing occurs at the SrTiO₃–NdNiO₃ interface and in the RP fault regions. Quantitative analysis of the variation in lattice constants indicates that large strains exist around the substrate–film interface. We demonstrate that the Ni valence change around RP faults is related to a strain and structure variation. This work provides insights into the microstructure and electronic-structure modifications around RP faults in nickelates.

1. Introduction

The recent discovery of nickel-based superconductors has filled the gap in nickel-based oxide materials in superconducting systems [1,2,3,4,5]. The infinite-layer phase NdNiO₂ can only be synthesized from the precursor ABO₃ perovskite structure by removing apical oxygen atoms from NiO₆ octahedra through soft-chemistry topotactical reduction [6]. For stabilization of the superconducting phase, chemical doping with divalent Sr replacing trivalent Nd is crucial. Optimal doping with the highest superconducting transition temperature (around 15 K) has been found in Nd_0.8Sr_0.2NiO₂ thin films [1]. During pulsed-laser deposition growth of Sr-doped thin NdNiO₃ films, a strong tendency to form Ruddlesden–Popper (RP) faults has been reported [2]. In addition, the epitaxial strain from the substrate can induce the formation of RP faults [7,8]. These RP faults display an atomic structure, where the inclusion of an additional AO layer breaks the long-range order of the ABO₃ perovskite phase [3,7,8,9,10,11,12,13,14,15,16,17]. As the microstructure of the entire film determines its electric properties, investigation of the defect structure and associated variations in the local electronic structure in nickelates is indispensable. Moreover, RP structures are related to a variety of physical and chemical properties, e.g., electro-catalytic activities [11], microwave dielectric performance [18], magnetic properties [19], and ferroelectric properties [20].

Epitaxial strain and/or a cation non-stoichiometry, e.g., an excess of A or B in the ABO₃ structure, can induce the generation of different RP faults that consist of an intergrowth of rock-salt-type and perovskite-type building blocks [11,12,21,22]. A single rock-salt layer can be described as an a/2 <111> or a/2 <110> stacking fault displaying a zigzag arrangement of cations [17]. The (AO)(ABO₃)_n RP structure forms when a rock-salt layer separates the n-layered structure of n(ABO₃) [9,23,24]. Detemple et al. [13] reported on three-dimensionally arranged RP faults in LaNiO₃–LaAlO₃ superlattices, where the authors observed an identical contrast of A and B columns in the ABO₃ structure in high-angle annular dark-field (HAADF) scanning transmission electron microscopy (STEM) investigations. Coll et al. [23] identified multiple configurations of RP faults in a LaNiO₃ film based on atomistic modeling, experimental HAADF-STEM investigations, and HAADF-image simulations. Bak et al. [11] demonstrated a strong tetragonal Jahn–Teller distortion of the RP-fault plane in LaNiO₃, proving a correlation between the octahedral distortion and the electro-catalytic performance of La-based perovskite oxides. However, to the best of our knowledge, influences of different RP-fault types on the variation in the electronic structure and especially in the Ni valence state have not been reported so far for NNO thin films.

Here, we characterize the local atomic structure of different RP-fault configurations and focus on the electronic properties around RP faults in the NNO film, based on advanced aberration-corrected scanning transmission electron microscopy (STEM) and electron energy-loss spectroscopy (EELS). RP-fault formation starts after a critical thickness of around two unit cells above the substrate–film interface, where different RP-fault types are found to be inhomogeneously distributed in the entire film. We identify an elemental intermixing in various RP-fault regions with different structural configurations. In addition, we demonstrate that the variation in the Ni oxidation state around different RP faults relates to the alteration of the chemical composition, possibly triggered by epitaxial strain.

2. Materials and Methods

The NdNiO₃ film was grown on a SrTiO₃ (001) single-crystal substrate at 630 °C in an ozone–oxygen atmosphere with a pressure of 1.8 × 10⁻⁵ Torr in an atomic layer-by-layer fashion in a custom-made ozone-assisted MBE setup. Then, it was cooled down after the growth in the same atmosphere. The fluxes of the effusion cells were calibrated prior to growth with a quartz crystal microbalance and monitored during growth by in situ reflection high-energy electron diffraction.

The TEM sample was prepared by automated tripod polishing (in an Allied MultiPrep System) until the sample was thinner than ~15 μm, after which Ar-ion milling at liquid N₂ temperature was performed by an ion polishing system (PIPS, Gatan, model 691) [25]. The milling step was performed at a voltage of 0.5 kV to reduce the damage to the surface of the sample. We used a spherical-aberration-corrected microscope (JEOL JEM-ARM200F, JEOL Co., Ltd., Tokyo, Japan) equipped with a DCOR probe corrector (CEOS GmbH, Heidelberg, Germany) and a Gatan GIF Quantum ERS with a Gatan K2 camera at 200 kV to acquire STEM images and EELS spectra. The convergence angle was set to 20.4 mrad and the collection semi-angles for high-angle annular dark-field (HAADF) imaging and annular bright-field (ABF) imaging were set to 70–300 mrad and 10–20 mrad, respectively. The EELS spectra were acquired at a dispersion of 0.5 eV per channel (with an energy resolution of ~1 eV). The principal component analysis (PCA) method [26] was utilized to optimize the signal and to reduce the noise in the EELS maps.

3. Results and Discussion

As the overview HAADF image in Figure 1a shows, the film thickness is around 9 nm, in good agreement with the X-ray reflectometry, and a large number of RP faults are distributed randomly throughout the entire film. The critical thickness for RP-fault formation is around two to three unit cells. In this range, the film grows coherently with the substrate. Similar observations were also made and are discussed in [2]. Above the critical thickness, defects are introduced to accommodate the strain. We note that this growth mode resembles the Stranski–Krastanov mode; however, it differs from it by not forming islands on the coherent NNO layer, but instead forming a continuous film. According to the relationship that the atomic-column brightness is approximately proportional to Z^1.7 (Z: atomic number), the sufficient contrast difference enables us to distinguish the atomic columns [27]. For example, we can observe a sharp interface between STO and NNO. A high density of irregularly distributed RP faults forms generally three unit cells away from the substrate.

Although we cannot exclude non-stoichiometry during film growth as a reason for RP-fault formation, a major driving force in our system is the reduction in the lattice mismatch between STO (3.905 Å) and NNO (pseudocubic lattice constant: 3.807 Å) as we do not observe any misfit dislocation that could alternatively have reduced the strain. In the enlarged HAADF image (Figure 1b), several configurations of RP faults in the NNO layer are marked with red dashed boxes (A, B, and C, respectively). The apparent a/2<110> shift for RP fault A, where a is the lattice constant of the pseudocubic unit cell, is likely to be the component of the true displacement vector a/2<111> projected along the beam direction [24]. RP fault B shows a single intergrowth layer of {NdNiO₃}, forming a typical ‘214’ structure of the RP phase (Nd₂NiO₄) by combining a rock-salt-type layer and a perovskite-type layer. In the case of RP fault C, the contrast difference between the A and B sites in NNO disappears due to the overlap of the A and B columns along the viewing direction. [13,23]. Additionally, the red dashed box D marks a fault-free region displaying a perovskite structure of the NNO layer.

Furthermore, we can observe the oxygen columns from the inverted ABF images (shaded in Figure 1) around the edges of these RP faults. The slight elongation of Ni–O bonds along the z-axis at the plane of RP fault A is apparent (see also Figure S2b), which can result in a significant splitting of the e_g level of the Ni 3d orbital and a relative variation in the O 2p level based on the Jahn–Teller effect [11]. From a layer-by-layer structural perspective [11], the distortion of oxygen anions toward the fault plane can reduce the strong repulsion between two [NdO]⁺ layers and achieve an energetically favorable structure. Around RP fault B, the elongation of Ni–O bonds is not clear, accompanying a visible distortion of [NiO₆] octahedra (see Figure S3). The reduced contrast of oxygen columns is possibly due to the existence of oxygen vacancies. Additionally, we observe distinct oxygen columns at RP fault C without apparent changes in the oxygen structure. Owing to the overlapping of elements around the edge of RP fault C, we observe a homogeneous distribution of oxygen.

To investigate the effects of RP faults on the electronic character of the NNO film, we performed STEM-EELS measurements with atomic resolution. Figure 2a,d show the HAADF images of RP faults A and B, respectively. For the sake of improving the signal-to-noise ratio, the O-K and Ni-L_2,3 edges were extracted from the red dashed boxes with a width of one unit cell shown in Figure 2b,e. From the NdNiO₃ reference, we observe a strong pre-peak A at the O K edge, which is associated with the hybridization of O 2p and Ni 3d orbitals, reflecting the covalency of Ni–O bonding and the oxidation state of Ni [28,29]. The pre-peak of the O K edge is weak or even absent around the RP faults, which is related to the changes in the local Ni–O distances due to epitaxial strain and RP faults [28,30]. In addition, the strong Nd–O interaction in the rock-salt layer results in a distinct shoulder of peak C. Close to the STO–NNO interface, the intermixing of Ti into Ni causes the appearance of peak B that arises from the Ti 3d–O 2p hybridized states. Peak B almost disappears in cells 4–7. Peak C corresponds to the O 2p states hybridized with Sr 4d in STO or Nd 5d in NNO. Peak D is due to the O 2p states hybridized with Ni 4 sp and Nd 5p states [31]. The L₃ and L₂ peaks of Ni arise from excitation of electrons from the 2p_3/2 and 2p_1/2 core levels to the unoccupied 3d band, which is used to determine the valence state of the transition metal from the white-line ratio and chemical shift [28]. The slight variations in peak positions and crystal-field splitting observed in Figure 2b,e imply changes in the electronic structure of Ni for both types of RP faults. We quantified the white-line ratio, which is shown in Figure 2c,f. From the DigitalMicrograph database [32], we calculated the Ni³⁺ and Ni²⁺ references, which are expected to correspond to the fault-free NdNiO₃ and NiO films. These are shown as black and blue dashed lines in Figure 2c,f, respectively. We observe that the L₃/L₂ ratio increases near RP-fault planes and in the fault-free NNO layers close to the STO substrate, suggesting a decreased valence of Ni. In the fault-free NNO layer (the seventh unit cell) far from the fault planes, the valence of Ni is close to 3+, as expected. The Ni L₃/L₂ intensity ratio in RP fault B is a little higher than that in RP fault A, indicating a lower valence of Ni around RP fault B.

The atomically resolved EELS maps shown in Figure 3 confirm the atomic arrangement around RP faults A and B. The distribution of cations from the color-coded EELS maps indicates elemental intermixing at the STO–NNO interface. In the RP fault regions, the zigzag arrangement of the Nd atoms and additional intergrowth of an NNO layer verify the atomic structures visible in ADF images. Figure 3a shows the EELS maps of RP fault A, where Ni and Nd columns can be clearly distinguished. In the RP fault B region (Figure 3b), a small amount of Ni is intermixed into Nd columns. Owing to the random distribution of RP faults, it is possible that the observed Nd/Ni signal intermixing arises from the partial overlapping of Ni and Nd. Similarly, the obvious intermixing of Ni and Nd signals at RP fault C shown in Figure S1a is due to the Ni/Nd overlapping, explaining why Ni and Nd columns display the same contrast in ADF images. Moreover, the Ni L_2,3 and O K edges extracted from fault-free and Ni/Nd overlapping regions show the same shapes and peak positions. The pronounced small pre-peak of the O K edge is consistent with the Ni³⁺ reference in Figure 2b,e.

In order to explore the potential mechanism of the valence change in Ni, we quantitatively analyzed the variation in the lattice constant around RP faults A and B. As the lattice constant of STO is a little larger than that of NNO, an in-plane tensile strain of ~2.5% is induced in the NNO film. From Figure 4a,b, the critical thickness of the NNO layer is two unit cells. By calculating the Nd–Nd distance across the STO–NNO interface and across the RP fault plane from the HAADF images, we found that the in-plane tensile strain is maintained at ~2.5% for the first two unit cells before the formation of RP faults. The in-plane Nd–Nd distance decreases to 3.85 ± 0.04 Å at the fault-free layers far from the RP fault plane shown in Figure 4c,d. According to theoretical calculations, the O stoichiometry in NNO is sensitive to the in-plane tensile strain [33]. The formation energy of an oxygen vacancy becomes negative when the in-plane tensile strain is higher than 1.9% [33]. Therefore, the probability for the generation of oxygen vacancies close to the interface is higher than that in fault-free layers where the in-plane strain is significantly released. In addition, oxygen vacancies are more easily generated in the NiO₂ plane as their formation energy in the NiO₂ plane is about 1.7 eV lower than in the NdO plane under a tensile strain [33]. From the inverted ABF image in Figure S2, we can clearly observe that the contrast of oxygen columns in the NiO₂ planes (marked with white arrows) is smaller than that in the NdO planes (marked with orange arrows), indicative of the possible existence of oxygen vacancies in our sample. Thus, the strain-induced oxygen vacancies in the region close to the STO–NNO interface can result in a decrease in the Ni valence. In addition, the sharp drop of the out-of-plane Nd–Nd distance to ~3.5 Å observed a few atomic columns away from the fault plane suggests a large out-of-plane contraction, which gradually increases to ~3.8 Å across the fault-plane and into the fault-free region as shown in Figure 4c,d. Therefore, the generation of RP faults effectively relaxes the epitaxial strain from the substrate, which is consistent with the previous report [12].

It is worth mentioning that the average oxidation state of nickel varies between 2+ and 3+ for different n in the RP structure (Nd_n+1Ni_nO_3n+1). In principle, Ni is divalent in the ‘214’ end member of the RP structure (Nd₂NiO₄), where the [NdO]⁺ layer donates an additional electron for the NNO layer [34]. However, the measured Ni valence in RP fault B shows a mixed valence between 2+ and 3+. The RP fault B structure is similar to the Nd₂NiO₄ structure, but the [NdO]⁺ layer is shared by adjacent NNO layers. We can observe the irregular distribution of different types of RP faults and fault-free NNO layers in Figure 1b. As mentioned above, at the NNO layer close to the substrate, the tensile strain can induce the formation of oxygen vacancies, which means that the additional electron can be partially compensated for by oxygen vacancies in the adjacent NNO layer. Thus, the Ni valence can be higher than 2+ in the intergrowth layer of NNO. Given that the Ni/Nd signal intermixing occurs at the intergrowth layer of NNO, it is possible that the measured Ni signal partially arises from the fault-free NNO layer, which can also lead to a mixed Ni valence in RP fault B. In terms of RP fault A, there is no obvious elemental intermixing and Ni ions show a higher valence (closer to 3+) than in RP fault B. The variation in the Ni oxidation state is mainly ascribed to the additional NdO layer and the possible oxygen vacancies.

4. Conclusions

Based on STEM and EELS measurements, we systematically studied the atomic structure of RP faults and the effects of RP faults on the local electronic configuration of Ni in a NdNiO₃ film. Several RP faults were detected, showing, e.g., a/2<111> shifts, intergrowth layers of NdNiO₃, and overlapping Nd and Ni columns. The critical thickness for RP fault formation is around two unit cells and no misfit dislocations are visible. The generation of a high density of RP faults effectively relaxes the epitaxial strain from the substrate. We observed the elongation and distortion of the oxygen sublattice around RP-fault planes, which is theoretically beneficial to reduce the strong repulsion between two neighboring [NdO]⁺ layers. We analyzed the variation in the Ni oxidation state around RP faults in NNO. The changed valence of Ni around RP-fault planes is mainly associated with the variations in the strain and atomic structure. On the one side, the tensile strain might induce oxygen vacancies at the fault-free layer of NdNiO₃ close to the substrate. On the other side, the additional [NdO]⁺ layer can lead to a reduction in the valence of Ni in RP fault regions. This work highlights how the distribution of strain and atomic structure changes around RP faults lead to a spatially varying electronic configuration in NdNiO₃.