Electronic structure of buried LaNiO₃ layers in (111)-oriented LaNiO₃/LaMnO₃ superlattices probed by soft x-ray ARPES

With the advancement of synthesis techniques, it is possible nowadays to control the growth of transition metal oxide (TMO) epitaxial heterostructures at the unit cell level. The realization of such heterostructures has resulted in the discovery of fascinating phenomena as a consequence of the confinement of strongly correlated electrons in ultrathin high quality layers.^1–3 Since these thin layers are embedded in heterostructures, interfacial phenomena such as charge transfer and magnetic coupling play an important role together with dimensionality in determining the ground state of the system.^4–8 The recent development of combined angle-resolved photoemission spectroscopy (ARPES) and oxide heterostructures’ growth setups has allowed the electronic structure of TMO heterostructures to be probed directly. However, the small electron escape depth in VUV-ARPES experiments has limited most studies to the top layer of oxide thin films.^2,9–11

Most members of the nickel based perovskite oxides RNiO₃ (R is a trivalent rare earth element) exhibit a metal-to-insulator transition as a function of temperature.^12–15 In addition, a transition from a paramagnetic to an antiferromagnetic ground state is observed in these compounds. The critical temperature for both transitions is a function of the Ni–O–Ni bond angle and can be controlled by changing the R ion size. The least distorted member of the nickelate family, LaNiO₃ (LNO), constitutes an exception since it remains rhombohedral, metallic, and paramagnetic at all temperatures. Despite presenting this rather simple bulk phase diagram, studies of thin LNO films have surged, motivated by the possibility of tailoring their properties by epitaxial strain,^16,17 thickness control,^18–20 and particularly by using them as a building block in epitaxial oxide heterostructures and devices.^3,21–29 It has been shown by transport experiments that while thick (001)-LNO films grown on different substrates are metallic, as the thickness is reduced, a metal to insulator transition is observed.^9,18 A similar dimensionality induced metal-insulator transition accompanied by the stabilization of an antiferromagnetic ground state was found in LAO/LNO superlattices.³ Photoemission spectroscopy experiments have consistently reported a loss of spectral weight at the Fermi level (E_F) accompanying the thickness driven metal-insulator transition; however, there is no consensus of the underlying physical mechanisms behind this observation.^9,30–32 ARPES experiments showed that the 3D electronic structure of (001)-oriented LNO thin films remains intact down to 15 Å, while for thinner films the destruction of Fermi liquid like quasiparticles⁹ and the appearance of 1D Fermi surface (FS) nesting³² were reported.

Interfacial and confinement effects in heterostructures grown along the perovskite-(111) orientation are much less studied.^24,29,33,34 Interestingly, in a thin film with a thickness of two monolayers (MLs) of LNO grown along this orientation, the Ni atoms form a buckled honeycomb lattice; such a heterostructure was predicted to host novel topological phases which are yet to be found experimentally.^35–38 The recent realization of high quality heterostructures that combine LNO with thin insulating ferromagnetic LaMnO₃ (LMO) layers along the (111) direction and exhibit unexpected phenomena is an example of the novel physics found in this type of heterostructures. Transport experiments reveal a metal to insulator transition in (111)-LNO/LMO superlattices as the LNO thickness is reduced. The magnetic properties of the insulating phase are rather surprising. When the samples are cooled down with an applied magnetic field, they display a negative exchange bias (EB).²⁴ The sign of the EB can be reversed as a function of temperature for heterostructures with an LNO thickness of 7 monolayers (MLs), defined as the distance between the consecutive Ni atoms of ∼2.19 Å. This behavior was interpreted based on the emergence of an AF spiral with a (1/4, 1/4, 1/4) wave vector in the nickelate layer that couples the ferromagnetic LMO layers.²⁸ 

Here we employ soft x-ray ARPES (SX-ARPES) to study the electronic structure of metallic (111)-LNO thin films for the first time. The choice of this experimental technique was dictated by the larger photoelectron mean free path and thus bulk sensitivity compared to the conventional VUV-ARPES as well as concomitant sharp definition of the surface-perpendicular momentum k_z.^39,40 The measured Fermi surface (FS) broadly agrees with previous experiments in (001) films and consists of an electron pocket at the Brillouin zone (BZ) center surrounded by eight hole pockets centered at the BZ corners.^9,10,41 We find that a simple two-band model of the Ni e_g states assuming a cubic BZ does not fully describe the electronic structure whereas density functional theory (DFT) calculations including the bulk octahedral tilt pattern closely resemble the experimental Fermi surfaces providing a solid base for our observations. The large photon energies used in our study allowed us to map the electronic structure of LNO layers embedded in LNO/LMO superlattices overcoming the inherent surface sensitivity of the technique and demonstrating its effectiveness in studying interfacial phenomena. Finally we show that concomitant with the thickness induced transition to insulating behavior in transport experiments, a loss of spectral weight at E_F is observed in heterostructures with 7 MLs thick LNO layers. However, no obvious sign of dimensional crossover or signatures of AF ordering are observed.

We have employed SX-ARPES to investigate the electronic structure of a LNO thin film and superlattices grown on STO-(111) oriented substrates. The experiments have been carried out at the SX-ARPES end station of the ADRESS beamline of the Swiss Light Source (Switzerland).³⁹ We used p-polarization of incident X-rays unless stated otherwise and the sample temperature was set to T = 12 K. All samples were grown ex situ by off-axis sputter deposition as described in Ref. 24 and exposed to air prior to the ARPES experiments. We first examined LNO₁₅, a thin film with a thickness of 15 MLs. Then, we proceeded to study the evolution of the electronic structure on two different [LNO_N/LMO_M]_X heterostructures composed of X repetitions of a LNO_N/LMO_M bilayer, where M and N are the number of monolayers of each material. In both cases the superlattices are terminated in a top LMO insulating layer. We explored [LNO₁₅/LMO₅]₄, where the LNO thickness in each layer matches the thin film, and then we probed [LNO₇/LMO₅]₆ a sample with reduced nickelate thickness expected to show EB and a spin spiral with ordering vector along the (1/4, 1/4, 1/4) direction. Resistivity measurements as a function of temperature are shown in Fig. 1(a) where two different behaviors are observed. A weakly metallic behavior in the sense that dR/dT > 0 is observed down to ≈150 K and 50 K for LNO₁₅ and [LNO₁₅/LMO₅]₄ samples. On the other hand, the sample with a thinner nickelate layer [LNO₇/LMO₅]₆ is clearly insulating (dR/dT < 0) consistent with the reported thickness induced metal-to-insulator transition. In Fig. 1(b) we show a magnetic hysteresis loop obtained at a temperature T = 5 K for the sample [LNO₇/LMO₅]₆ under zero field cooled (blue) and 0.05 T field-cooled (red) conditions. The 190 Oe shift in the field cooled hysteresis loop confirms the presence of EB in the sample, while this effect is not observed for [LNO₁₅/LMO₅]₄ with a thicker LNO layer (not shown).^24,28

The results of our electronic structure measurements by means of SX-ARPES performed on LNO₁₅ are shown in Fig. 2. Fermi Surface (FS) measurements obtained with 500 and 382 eV excitation energy are displayed in Figs. 2(a) and 2(b). These energies correspond to the high symmetry points Γ and R in normal emission for the cubic unit cell considered here (see supplementary material).³⁹ In Fig. 2(j) we present a tight binding calculation of the bulk FS which consists of an electron pocket centered at Γ and hole pockets centered at the R points. The calculations assume a ratio between nearest-neighbor hopping (t) and next-nearest-neighbor hopping (t’) t’/t = 0.13 known to be appropriate for describing the (001)-LNO FS.⁴² In Figs. 2(d) and 2(e) we show the corresponding cuts of the tight-binding calculation through the high symmetry points in order to better compare with the experimental data. Electron pockets are shown in red and hole pockets in green. The experimental FS centered at Γ is formed by a central quasi-hexagonal electron pocket surrounded by triangular hole pockets. These features are in fair agreement with the calculations. Dispersion plots measured along k_{[1 1 $2¯$]} direction obtained at different Γ points are shown in Figs. 2(k) and 2(l). We observe that the band derived from the e_g states of Ni which forms the electron pocket extends up to ∼170 meV. A similar bandwidth was observed for thin films grown along the (001) direction indicating that distortions imposed by the substrate do not play a major part in determining this parameter.⁴³ We then turn our attention to the k-space cut through the R point at the BZ corner. In this measurement shown in Fig. 2(b), the electron pocket is absent as expected while large Fermi surface contours corresponding to cuts through the center of the bulk hole pockets are found at the locations predicted by the model in Fig. 2(e). However the correspondence between calculation and data is not entirely satisfactory. The small hole contours predicted by the model are not resolved clearly in the experiment and the shape of the large hole contours differs slightly from the model. These subtle discrepancies could arise from the use of a simplified cubic unit cell in our tight binding model and/or from more intricate structural or electronic effects at the LNO/LMO interfaces. In order to distinguish these scenarios, we compare in Figs. 2(g) and 2(h) our tight-binding model to a bulk DFT calculation that includes a trigonal distortion due to strain in the film and the bulk rotations of the NiO₆ octahedron. We first notice that the DFT-FS obtained for the Γ and R points is identical. This is due to the presence of backfolded bands, which are generally weak in ARPES (see supplementary material). In order to highlight the fundamental bands, we overlay the tight-binding model with a single atom in the basis. Interestingly we see that the hexagonal-like central hole pocket in Fig. 2(h) is rotated 30° in the DFT calculation with respect to the tight binding model. This is an effect of the octahedral rotations and in fact is observed in the data where we see that the sides of the hexagonal-like pockets are facing each other showing that the bulk-like rotations of LNO are not fully suppressed by the STO substrate.

In order to complete our characterization of the electronic structure of LNO₁₅, we scanned the excitation energy in the 350–650 eV range to study the k_z dependence of the electronic structure. The results shown in Fig. 2(c) reveal a strong dispersion along k_z, characteristic of highly 3D electronic states. As is evident from a comparison with the model shown in Figs. 2(f) and 2(i), only half of the electronic bands are detected when measuring in this configuration. We confirmed that the bands are neither observed when measuring with s-polarized light suggesting an intricate matrix element effect. Taken together these measurements confirm the 3D nature of the electronic structure of the LNO₁₅ schematized by the tight binding model depicted in Fig. 2(h), with the caveat that the shape of the hole pockets can only be correctly described if the bulk-like rotations of the NiO₆ octahedron are taken into account as in the DFT calculations shown in Fig. 2(h).

In Figs. 3(a) and 3(b) we show constant energy maps measured on the LMO terminated [LNO₁₅/LMO₅]₄ and [LNO₇/LMO₅]₆ heterostructures, respectively, obtained with 500 eV excitation energy. The electron inelastic mean free path for this energy is ∼15 Å which limits the probing depth of our study to the whole top LNO_N/LMO_M bilayer.⁴⁴ The observed electronic structure is markedly similar for the [LNO₁₅/LMO₅]₄ sample and the LNO₁₅ film (cf. Fig. 2(a)). Even though the LNO thickness is the same in both layers, one would expect a modified FS as a consequence of the magnetic coupling and charge transfer at the interfaces.⁴⁵ However, the only appreciable effect in our data is a slight widening of the features in the spectra of the heterostructure, possibly due to the finite roughness at the interface with the LMO top layer. When the thickness of the LNO layer is further reduced down to 7 MLs, there is a considerable reduction of spectral weight at the Fermi level consistent with the insulating character of the sample and the band-like features broaden significantly. Intriguingly though, the k-space map obtained by integrating spectra over a narrow window around E - E_F = 50 meV does not show a marked redistribution of spectral weight as it is a characteristic for nesting driven itinerant spin-density wave phases.^46,47 While these findings do not exclude AF ordering in LNO as it was invoked in Ref. 28 to explain the exchange bias observed in superlattices with 7 MLs LNO,^24,28 they suggest that the insulating behavior of thin LNO layers in LNO/LMO superlattices is triggered by a gradual loss of quasiparticle coherence, rather than a well-defined phase transition to a magnetically ordered state.

We further investigated the loss of spectral weight at the Fermi level by analyzing the angle integrated spectra of the three mentioned heterostructures. In the valence band spectra of the superlattices shown in Fig. 4(a), there is a distinct peak ∼2.3 eV below the Fermi level that is absent in the thin films and corresponds to the top LMO layer. In the near-E_F region, we identify two peaks located ∼0.1 and 1 eV below E_F. The optical gap of LMO has been reported between 0.6 and 1 eV and previous XPS measurements revealed the onset of the density of states at 0.5 eV below the Fermi level.^48–51 Hence, we conclude that the peak located 0.1 eV below E_F corresponds to e_g states in Ni with the feature at 1 eV likely having contributions from both LMO and LNO. Figure 4(b) depicts the region of the spectra close to the Fermi level where the loss of spectral weight at E_F for the [LNO₇/LMO₅]₆ sample is evident. In order to quantify this effect, we show in Fig. 4(c) empirical fits using a spectral function consisting of a Lorentz shaped peak multiplied by a Fermi function and a Gaussian to take into account the instrument resolution. The best fit to the integrated spectra shown in the figure is obtained by considering an effective Fermi level at -50 meV. Similar fits performed for k-resolved spectra throughout the BZ do not show a significant variation of this energy scale. Again, this is untypical for a nesting driven instability of the Fermi surface where the gap size would be expected to correlate with the degree of nesting at the ordering wave vector.⁵² Further considering the absence of a large redistribution of spectral weight in momentum space discussed previously, we thus interpret the suppressed low-energy spectral weight as a pseudogap in the sense that is not associated with a clear symmetry breaking. Such a pseudogap is fully consistent with the insulating character observed in transport and might arise from strong local correlations, possibly in conjunction with disorder due to interface roughness.

In summary, we reported the ARPES Fermi surface of a LaNiO₃ thin film grown along the less-conventional (111) crystallographic direction. The observed 3D-FS consisting of an electron pocket at $Γ$ and hole pockets centered at the R points of the BZ is qualitatively consistent with bulk models and previous observations in (001) films. When incorporated into [LNO/LMO] heterostructures, the gross electronic structure of LNO changes only slightly with the spectral weight distribution conserving a 3D character. For the thinnest LNO layers, our ARPES data show a loss of weight at the Fermi level indicative of a pseudogap opening and consistent with the observation of an insulating behavior in transport experiments.

See supplementary material for a full description of DFT calculations, a discussion of k_z determination, and figures showing the difference between Fermi surfaces determined by spherical and plane k-space contours.

This work was supported by the Swiss National Science Foundation through Division II (Nos. 200021_146995 and 200021_146890), Ambizione (Grant No. PZ00P2_161327), and NCCR MARVEL. The research leading to these results has received funding from the European Research Council under European Union’s Seventh Framework Program (No. FP7/2007-2013)/ERC Grant Agreement No. 319286 (Q-MAC) and No. 290605 (PSI Fellow/Cofound).