Graded orbital occupation near interfaces in a ${\mathrm{La}}_2{\mathrm{NiO}}_4-{\mathrm{La}}_2{\mathrm{CuO}}_4$ superlattice

X-ray absorption spectroscopy and resonant soft x-ray reflectivity show a nonuniform distribution of oxygen holes in a ${\mathrm{La}}_2{\mathrm{NiO}}_4-{\mathrm{La}}_2{\mathrm{CuO}}_4$ (LNO-LCO) superlattice, with excess holes concentrated in the LNO layers. Weak ferromagnetism with $T_c=160$ K suggests a coordinated tilting of ${\mathrm{NiO}}_6$ octahedra, similar to that of bulk LNO. Ni $d_{3z^2-r^2}$ orbitals within the LNO layers have a spatially variable occupation. This variation of the Ni valence near LNO-LCO interfaces is observed with resonant soft x-ray reflectivity at the Ni and Cu L edges, at a reflection suppressed by the symmetry of the structure, and is possible through graded doping with holes, due to oxygen interstitials taken up preferentially by inner LNO layers. Since the density of oxygen atoms in the structure can be smoothly varied with standard procedures, this orbital occupation, robust up to at least 280 K, is tunable.

Figure 1

(a) STEM high-angle annular dark field image of a LNO-LCO superlattice showing epitaxial growth. The right panel expands the image area enclosed by the rectangle. The bright features in the image are La atoms in the LaO planes, represented by dashed lines between the $\mathrm{TM}-{\mathrm{O}}_6$ octahedra. The oxygen atoms are not visible. The experimental conditions were $1.3\AA$ probe size at 200 kV, $28\mathrm{mrad}$ convergence angle, and $64–341\mathrm{mrad}$ collection angle. (b) Wide-angle hard x-ray measurements made at Advanced Photon Source at Argonne National Laboratory (7.5 keV) and with an X’ Pert diffractometer (8.05 keV) showing superlattice peaks (SL) near the main reflections (M) indexed with respect to the measured superperiod. Substrate peaks (STO) are also visible.

Figure 2

$L$ scans of the low momentum range for hard and soft x-ray scattering. HXR measurements show strong thickness oscillations, with a peak at $L=1$ and a step at $L=2$ (black line) for ID6 data. The peaks at $L=3.8$, 4.6, and 5.7 are stray reflections. Soft x-ray scattering shows pronounced superlattice (SL) peaks for $L=1$ at O K, Ni L, and Cu L edges. The Ni $L=1$ peak is shifted because of refraction. The $L=2$ reflection is much larger at the Ni edge compared to the Cu edge. This observation is confirmed by detailed energy-resolved measurements (Fig. 6), which are discussed below. In contrast, the absence of the $L=2$ reflection in measurements at the oxygen edge is caused by the energy used (see Fig. 5 for a complete measurement). Smaller differences can be observed between the measurements at the Cu and Ni edges, for instance, a slight indication of a peak at $L=4$ at the Cu edge. The small difference in intensity at $L=4$ can be explained using the different polarizations of Ni and Cu $d$ orbitals, unlike the opposite and larger difference at $L=2$ (Sec. 2e). The intensity of SL reflections at $L=3$ and 4 is small, consistent with the calculated dependence of SL structure factors amplitude on $L$. The $L=5$ and 6 SL peaks are larger because of the proximity to the Bragg peak corresponding to 1 ML (off $x$ axis in this figure).

Figure 3

Magnetic moment measurements show a ferromagnetic transition at 160 K. A diamagnetic contribution has been subtracted from the data. The applied magnetic field was 100 Oe. Resistivity measurements show an insulating behavior down to 38 K with a small anomaly in slope at 58 K.

Figure 4

(a) Electron and fluorescence yield and the correspondence of the peaks to the superlattice layers. The LCO UHB peak disappeared completely for an LNO-SCO superlattice grown on an STO substrate, showing that there is no STO substrate or LNO layer contribution at the LCO UHB energy. From detailed measurements (not shown) on LNO-SCO superlattices LNO MCP and LCO MCP have different energies. (b) RSXS intensity (logarithmic scale) at 90 K at the O edge shows two peaks corresponding to the first two peaks in TFY, with the lower (529 eV) and higher (531.5 eV) energy peaks from the LNO and LCO layers, respectively. The different $L$ profiles are due to an interference effect and different layer $c$ axis parameters.

Figure 5

Intensity of the $L=2$ reflection at the O edge at 90 K (logarithmic scale). The three peaks are (in order of increasing energy): LNO holes at $E=529.5$ eV, LCO UHB holes at $E=531.5$ eV, and a feature at $E=533.5$ eV, discussed briefly in Sec. 3b. Unlike the scattering at $L=1$, the LNO MCP peak has substantially smaller intensity than the LCO UHB peak.

Figure 6

(a) Scattering intensity (logarithmic scale) at La ${\mathrm{M}}_4$ ($E=853.25$ eV) and Ni L${}_3$ ($E=856$ eV) edges at $L=2$. The increase in the intensity over background at the Ni edge is 100$\%$. The features below 853 eV are thickness oscillations. No off-resonant intensity tail is observed near $L=2$ at the Ni edge, showing that this reflection originates in an electronic state and not from, for instance, selective removal of Ni ions from the LNO layers. (b) Scattering intensity (logarithmic scale) over a wide range at the Cu edge near $L=2$ for comparison. The energy range (14 eV) is chosen the same as in (a). (c) The area of interest at the Cu edge magnified, compared to TFY and TEY. A very small peak is visible for $L=1.96$, with a relative change in intensity over background of only 6$\%$. Because of the large sloping background an $L$ scan is less useful in showing this small peak.

Figure 7

(a) $L$ scans at the Ni ${\mathrm{L}}_3$ edge show the $L=2$ peak for sample azimuthal orientations with incident beam polarization at ${45}^{\circ }$ and along the tetragonal $a-b$ axes. (b) The $L=2$ peak at the Ni ${\mathrm{L}}_3$ edge persists up to at least 280 K. As will become clear in Sec. III.A, this is due to the larger energy necessary to move the oxygen dopants in the structure compared to the thermal energy needed to destroy a spin order. (c) The $L=2$ peak at the Ni ${\mathrm{L}}_2$ edge. The inset shows the energy profile of this peak (circles) compared to TFY (squares) and TEY (blue line).

Figure 8

(a) Calculated XAS (offset vertically for clarity) using the software “Missing”[30] for the ${\mathrm{Ni}}^{3+}$ and ${\mathrm{Ni}}^{2+}$ valences are in good agreement with Ref.29. The parameters used were $10Dq=1\mathrm{eV}$, $Ds=Dt=0$, 0.8 Slater integrals scaling, 0.2 eV Lorentzian, and 0.1 eV Gaussian broadenings. Instead of the full tetragonal $D_{4h}$ symmetry, which further splits the final $d$ states, a cubic crystal field, with $d$ states of octahedral $O_h$ symmetry, has been used for simplicity. The lower panel shows the measured electron yield (navy line), fluorescence yield (red squares) at La ${\mathrm{M}}_4$ and Ni ${\mathrm{L}}_3$ edges, and the fit of TFY (red line) with three peaks at the La and Ni edges (green) on a smooth step function (blue). The scattering at $L=2$ (black squares) peaks at the first Ni peak only. The lowest-energy peak at $851\mathrm{eV}$ is a fringe effect, seen more clearly in Fig. 6a. (b) Sketch of the layer-resolved orbital occupation where, for clarity, a LNO layer with 4 ML has been illustrated, and only the $d_{3z^2-r^2}$ orbitals have been included. Dashed lines represent the double LaO layers between which oxygen interstitials are located. Their number has been exaggerated in this sketch and the trilobe drawing (not to scale) illustrates their hybridization with La ions in the two adjacent LaO layers along axes at ${90}^{\circ }$ with each other.[15] A small spin gradient is also shown at the interface.