Phase diagram of compressively strained nickelate thin films

The rich phase diagram of many strongly correlated oxides offers the possibility of creating devices and heterostructures displaying novel functionality. Such devices can exploit sharp metal-insulator transitions, for instance, for use in resistive switches and phase-change memories. Indeed, the idea of a Mott transition field effect transistor, or Mott-FET, has been considered as a potential alternative to conventional Si-based technologies because of the absence of extrinsic limitations – such as short channel effects – due to the high carrier densities characteristic of these oxides.^1,2 In this regard, the rare-earth nickelate perovskites (RNiO₃) comprise an ideal model system, as the influence of the electronic correlations on phase behavior can be tuned by the choice of rare earth cation (R), temperature, dimensionality, and applied stress.³ 

The metal-insulator transition in RNiO₃ is driven by interactions between Ni-3d and O-2p electrons which are sensitive to changes in structure and temperature.³ The transition may be modified by three forms of stress. The first is intrinsic, arising from steric limitations of the RNiO₃ distorted perovskite structure. The ideal perovskite structure consists of |${\rm NiO}_6^{3 - }$|$NiO 63−$ octahedra linked at their corners with R³⁺ ions in the interstices. The R³⁺ ion is relatively too small to fit this ideal structure, exerting an internal stress that is accommodated by a rigid rotation of the |${\rm NiO}_6^{3 - }$|$NiO 63−$ octahedra. This rotation decreases the Ni–O–Ni bond angle away from the ideal value of 180°, narrowing the bands. The metal-insulator transition temperature, T_MI, thus decreases with increasing R³⁺ ion size, varying from 570 K for HoNiO₃ to 130 K for PrNiO₃·La³⁺ is so large that LaNiO₃ remains metallic to zero temperature. The transition temperature can be tuned continuously using solid solutions of rare earths.^4,5 A second form of stress, external hydrostatic pressure, decreases the rotation of the octahedra, thereby decreasing T_MI.^6–8 

Thin films of RNiO₃ experience a third form of stress due to mismatch of the lattice and thermal expansion to the substrate, in addition to the intrinsic internal stress. Epitaxial growth of pure NdNiO₃ films on a variety of substrates has shown a renormalization of T_MI with respect to the bulk, that is, T_MI is increased (decreased) by tensile (compressive) in-plane strain, an effect which is attributed to an increase (decrease) of the Ni–O–Ni, bond angle which affects films in the same way as the bulk.^9–14 Alternatively, it has been suggested that tensile in-plane stress may lead to a novel breathing distortion that creates two inequivalent Ni sites; charge transfer between these sites increases T_MI.¹⁵ 

Comparison of films is complicated by misfit dislocations, which partially relieve epitaxial strain and create strain gradients through the film thickness. Misfit strain can be controlled by growing defect-free films in registry (i.e., without misfit dislocations) with a low-misfit substrate, such as LaAlO₃. It has been shown that, in such films, rotation of the |${\rm NiO}_6^{3 - }$|$NiO 63−$ octahedra accommodates about half the epitaxial strain; the other half leads to octahedral distortions (i.e., elongation along the c-axis.).¹⁶ 

The strong effects of epitaxial strain across the RNiO₃ phase diagram should be understood to enable novel device integration and to further the fundamental understanding of the relation between atomic structure and electronic properties. Previous work using the different misfits of various substrates to tune the transition has suffered from inconsistencies in the growth.^9,14 Systematic results may be obtained by tuning the rare-earth composition with Nd_1-yLa_yNiO₃ (0 ≤ y ≤ 1) solid solutions while maintaining the same growth conditions. In this work, we study such alloyed films grown in registry with LaAlO₃(001) substrates (pseudocubic indexing is used throughout), providing low-defect films with a homogeneous compressive strain; in addition, we use sample preparation aimed to avoid oxygen vacancies.⁹ We characterize the transport, structural, and electronic properties of the films using resistivity measurements, synchrotron diffraction crystal truncation rod data, and x-ray absorption spectroscopy. From this, we produce an electronic phase diagram for thin films resembling that of the bulk, but with lower transition temperatures (renormalization) due to the effect of epitaxial strain.

The Nd_1-yLa_yNiO₃ (0 ≤ y ≤ 1) films in this study are grown on (001)-oriented LaAlO₃ substrates using molecular beam epitaxy (MBE) assisted by a radio-frequency oxygen plasma source. The bulk pseudocubic lattice parameters for NdNiO₃, LaNiO₃, and LaAlO₃ are approximately 3.81 Å, 3.84 Å, and 3.79 Å, respectively. These values translate to a lattice mismatch of 0.5% for NdNiO₃ and 1.3% for LaNiO₃. LaAlO₃ provides a nearly cubic template, with lattice planes rotated by <0.1° from the perovskite structure and twin boundaries typically >100 μm apart. The thicknesses of the films are held to 8 unit cells (uc). The Nd, La, and Ni metals are co-deposited in an oxygen environment with an oxygen partial pressure of 5 × 10⁻⁶ Torr. The temperature of the substrate during growth is 590 °C for each film. We utilize a quartz crystal microbalance to determine the relative stoichiometry of the metals prior to growth and ensure a growth rate of 1 layer/min. During the deposition, we monitor the film surface using reflection high energy electron diffraction (RHEED). The specular reflection shows intensity oscillations consistent with a layer-by-layer growth mode, with the peak-to-peak time corresponding to the growth of one unit cell. The observed sharp post-growth RHEED patterns along the [100] and [110] directions indicate coherent epitaxy.²⁶ To eliminate the presence of oxygen vacancies from our films, the samples are annealed in flowing O₂ for 6 hours at 600 °C following growth.

For each film, we measure the resistivity as a function of temperature (ρ-T) using a conventional van der Pauw geometry with Au contacts on the corners. We place the samples in a Quantum Design ⁴He cryostat to control the temperature in the range 10–300 K. Figure 1 shows the ρ-T curves obtained for 8 uc films with y = 0 (pure NdNiO₃), y = 0.1 (90% Nd/10% La alloy), y = 0.25 (75% Nd/25% La alloy), and y = 0.4 (60% Nd/40% La alloy). We define the metal-insulator transition temperature, T_MI, as the temperature at which ρ reaches a minimum.¹⁷ The curves for y = 0, 0.1, and 0.25 show metal-insulator transitions with T_MI ≈ 155 K, 129 K, and 68 K, respectively. For y = 0.4, the film remains metallic in the entire temperature range. The high temperature metallic phase is characterized by a decrease in resistivity as the temperature is decreased (dρ/dT > 0). In the insulating state, the temperature dependence of the resistivity changes such that dρ/dT < 0. The samples with insulating ground states exhibit a temperature hysteresis in ρ around T_MI upon cooling and heating the samples. The width of the maximum hysteresis window is 3 K for y = 0 and 10 K for y = 0.1 and 0.25, which also closely matches the difference in T_MI upon heating and cooling. For clarity and for comparison with bulk measurements, Figure 1 shows only the data obtained upon heating.

To map out a diagram of electronic phases as a function of the La content, y, we determine T_MI from the ρ-T measurements for 8 uc Nd_1-yLa_yNiO₃ alloys with y = 0, 0.1, 0.25, 0.4, 0.5, 0.9, 1. The resultant phase diagram is shown in Figure 2. For y ≥ 0.4, all films remain metallic down to T < 10 K, and thus we use a value of T_MI = 0 for these films. The (y, T_MI) points for y = 0, 0.1, and 0.25 are taken from the data shown in Figure 1. In analogy with the corresponding bulk phase diagram from Ref. 5, we also plot the transition temperatures as a function of the average size of the rare-earth ion (12-fold coordinated R³⁺ ionic radii as determined by Shannon¹⁸). We find that the thin film and bulk diagrams share a similar y-T_MI dependence, despite the clamping effect of the substrate and the nearly two-dimensional thickness of the films.⁵ The value of T_MI at y = 0, however, is ∼50 K lower for the NdNiO₃ (bulk T_MI ≈ 195 K), indicating that the MI transition temperature is shifted to lower temperatures for thin films. Furthermore, we find that a coherently strained 48 uc bulk-like NdNiO₃ film grown on LaAlO₃ has a T_MI ≈ 156 K, matching that of the 8 uc NdNiO₃. Thus, heteroepitaxial strain, rather than the reduction in dimensionality, appears to be the most likely source of the T_MI shift.

In order to understand the origin of the similarities and differences between the bulk and thin film phase diagrams more thoroughly, we examine the structural properties of the thin film series using synchrotron-based x-ray diffraction. We carry out crystal truncation rod (CTR) measurements at the Advanced Photon Source at XOR beamline 33 ID-D at an x-ray energy of 16 keV. Figure 3(a) shows a comparison of the CTRs with H = 2 and K = 0 measured for 8 uc Nd_1-yLa_yNiO₃ films, with y = 0, 0.25, and 1, along with their associated fits.²⁶ The out-of-plane lattice constants, c, of the films are determined from fits to the CTR data and compared for the three films in Figure 3(b). We observe a linear increase in c (Figure 3(b)) and the unit cell volume (Figure 3(c)) with increasing substitution of La. However, the unit cell volume of the thin films is reduced relative to their corresponding bulk analogs, as shown in Figure 3(c). This result indicates that the external strain caused by the substrate-film epitaxial mismatch leads to a reduction in bond lengths and octahedral tilt angles, as in the case of applied hydrostatic pressure.¹⁹ 

As mentioned earlier, this type of strain effect and the resultant change in the transition temperature has been observed in previous studies of pure NdNiO₃ films,^{11,14,20–22} but the observed renormalization across the entire phase diagram – the rigid shift to lower T_MI – demonstrates the systematic effect of compressive epitaxial strain on ultrathin films for all compositions studied here (between pure NdNiO₃ and LaNiO₃). It should be noted that other work on thin NdNiO₃ films grown on LaAlO₃, however, have determined different values of T_MI than that reported here, ranging from 175 K to 0 K.^12,13 This difference may be explained by divergent growth conditions, as the electrical properties of RNiO₃ films vary with oxygen content, which can depend sensitively on sample preparation.^23,24 Further work is needed to determine what growth factors are most crucial to explain these changes. In this study, we ensure oxygen stoichiometry by using atomic oxygen during growth as well as post-growth annealing. Despite growth disparities, a renormalization for compressively strained NdNiO₃ has been consistently observed, thus one can expect that the qualitative structure of the thin film phase diagram of Figure 1 is independent of growth conditions.

In addition to examining the physical structure, we determine the effects of alloying on the electronic structure of the films by performing soft x-ray absorption (XAS) measurements at the U4B beamline at the National Synchrotron Light Source (NSLS) at room temperature. We measured spectra for the Nd_1-yLa_yNiO₃ films at the O-K edge and the Ni-L₂ edge at room-temperature (the x-rays are linearly polarized perpendicular to the c axis; these films show no significant linear dichroism). As a reference, we also measure a bulk LaAlO₃ crystal and a 48 uc bulk-like NdNiO₃ film. The pre-peak regions of the measured O-K edge spectra are shown in Figure 4. For comparison, the data are scaled to the number of Ni layers using the integrated area under the measured Ni-L₂ peak, which indicates a nominal Ni³⁺ oxidation state for each film. The pre-peak at ∼531 eV is absent for the LaAlO₃ substrate and is associated with hybridization of the Ni-3d and O-2p states that govern conduction in the system.²⁵ As seen in the figure, the 48 uc and 8 uc NdNiO₃ films have nominally identical normalized pre-peak intensities, indicating the compositional origin of this feature. We observe an increase in the peak intensity with La content for our films. Accordingly, we attribute the increasing conductivity we find with the addition of La to an increase in the Ni-3d–O-2p hybridization.

In summary, we have constructed a phase diagram for 8 uc nickelate films under compressive epitaxial strain. This phase diagram is qualitatively similar to the bulk phase diagram, but with an overall renormalization of the metal-insulator transition temperature, T_MI. The reduction of T_MI with respect to the bulk can be explained by the modification of the structure caused by substrate clamping. In addition, the observed change in T_MI as a function of La content arises from an increase in the overlap between Ni-3d and O-2p orbitals, as has been found in the bulk due to structural modifications induced by strain and changes in the effective rare-earth ionic radii. The conclusions drawn here may be generically applicable to other strongly correlated oxide thin films experiencing metal-insulator transitions and useful for the future utilization of such materials for novel device applications.

This work was supported by DARPA Grant No. W911NF-10-1-0206 and NSF MRSEC DMR 1119826 (CRISP). Use of the Advanced Photon Source is supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences under Contract No. DE-AC02-06CH11357. Use of the National Synchrotron Light Source, Brookhaven National Laboratory, is supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. DE-AC02-98CH10886.