Near-nanoscale-resolved energy band structure of LaNiO₃/La_2/3Sr_1/3MnO₃/SrTiO₃ heterostructures and their interfaces

Complex oxide interfaces have been of interest for their many exciting properties and the physical mechanisms underlying them. Thus, heterojunctions between two such insulating oxides can exhibit superconducting features;¹ likewise, the interface between two antiferromagnetic complex oxides can become ferromagnetic.^2,3 A key goal of complex oxide research is the creation and control of spin states in heterojunctions and superlattices using charge transfer, proximity effects, and locally broken symmetries. One such complex oxide of particular interest is LaNiO₃ (LNO), a nickleate widely studied for its metal/insulator properties.⁴ Likewise, La_2/3Sr_1/3MnO₃ (LSMO) is a complex oxide studied extensively for its magnetic properties.⁵ Confinement effects at LaNiO₃ superlattices with other complex oxides are known to produce metal-insulator transitions,⁶ magnetization,⁷ and possibly superconducting behavior.⁸ LNO/LSMO superlattices are of high interest for their enhanced magnetic charge state produced by charge exchange across their interfaces.⁹ The local electronic structure of these interfaces depends significantly on their band alignment and dipoles introduced by defects during growth as a result of lattice mismatch, strain-induced distortion, and interdiffusion.

Depth-resolved cathodoluminescence spectroscopy (DRCLS) is an effective technique to measure the density distribution of native point defects and the band structure in a wide range of semiconductors and insulators.^10,11 This technique allows for measurement of optical transitions between filled and occupied energy bands below and above the Fermi level E_F as well as transitions at energies below the fundamental band gap involving lattice defects. We used a Monte Carlo simulation^12,13 at low (0.5–5 keV) incident electron beam energies E_B, to obtain depth profiles of electron-hole pair creation rates. This program simulates complete trajectories for each electron in an incident electron beam of energy E_B that scatters within a solid of specified density, weight fraction, and atomic weight of the elements comprising the solid. The profiles generated for this study provided the maximum penetration depth in the sample of the electron trajectories Z_Max including backscattered electrons Z_{MaxBackscattered} (see the Appendix). The incident electron beam generates a cascade of secondary electrons that lose energy initially by generating x-rays, then plasmons at lower energies, and finally impact ionization, i.e., electron-hole (e-h) pair creation, and phonons before coming to rest.¹⁴ Since the scattering length of electrons with energies below plasmon energies of semiconductors and insulators is less than 1–2 nm,¹⁵ the Z_Max and Z_{MaxBackscattered} profiles together effectively provide a profile of the e-h pair creation rate with this precision. The peak rates U₀ and maximum range, e.g., Bohr–Bethe range R_B for E_B in the 0.5–5 keV range can be localized in each layer of the 30 nm LNO/30 nm LSMO/STO heterostucture. Minority carrier lifetimes and diffusion lengths are short (nanometers to tens of nanometers) inside such structures due to high densities of recombination centers in these oxides.¹⁶ Short diffusion lengths are also found for quantum wells and other confinement structures that limit minority carrier diffusion. The depth resolution is enhanced further using a differential form of DRCLS (DDRCLS) so that we can probe each of these layers individually and extract information about the interior of each layer and their interfaces. Previous studies of ultrathin interfacial species provide evidence of this depth resolution.^17,18 Here, we demonstrate the ability of DDRCLS to measure optical transitions involving the electronic energy levels in LNO and LSMO films and their interfaces with this enhanced depth resolution. We detect optical transitions between orbital-derived energy levels with filled states near the Fermi level. Additionally, we detect changes in the electronic structure at the interfaces on a near-nanometer scale.

Epitaxial layers of LMO, LSMO, and LNO were grown coherently on (001)-oriented SrTiO₃ (STO) at 600 °C using oxide molecular beam epitaxy. The growth was monitored in situ by reflection high-energy electron diffraction and oscillations in the specular spot intensity corresponding to deposition of a single unit cell were observed, indicating a layer-by-layer growth mode. Heterojunction samples consisted of 30 nm La_2/3Sr_1/3MnO₃ (c = 0.387 nm) followed by 30 nm LaNiO₃ (c = 0.382) on (001) STO. These thicknesses were similar to those of LNO/La_2/3Sr_1/3MnO₃ (Ref. 19) and LNO/LMO (Ref. 20) superlattices and enabled DRCLS measurements with high depth resolution of the individual layers as well as their interfaces. A thicker (60 nm) LNO on STO sample provided additional confirmation of the bulk-like electronic transitions of the epilayers. X-ray reflectivity, high-resolution x-ray diffraction (XRD), and Z-contrast scanning transmission electron microscopy measurements on LNO/LSMO superlattices grown previously under identical experimental conditions and thicknesses provide evidence that the LNO/LSMO interfacial roughness is less than one unit cell.²⁰ 

Polycrystalline LSMO powder was produced by grounding SrCO₃, MO₂, and anhydrous La₂O₃ in a mortar_. The mixture was ramped to 1200 °C at a rate of 10 °C per minute and then left to anneal for 72 h. The materials diffuse to create LSMO as a final product. The sample temperature was then decreased at the same ramp rate. XRD was used to confirm the crystallinity of the powder.

We obtained DRCLS spectra using a glancing incidence electron gun, an Oriel monochromator, and CCD detector in a UHV chamber with electron beam energies E_B = 1.0–5 keV employed in 0.5 keV increments. Incident electron beam angle was 45° with 0.2–2 μA currents for constant power. The spot size of the beam was approximately 300 μm. Spectral resolution was ≤0.1 eV. No effects of sample charging were observed. A closed cycle He cold finger and braid connected to the sample holder enabled measurements at both room temperature and ∼80 K.

Figure 1 illustrates 45° incident Monte Carlo distributions for the rate of electron-hole (e-h) pair creation at STO/30 nm LSMO/30 nm LNO heterostructures. We used DRCLS to obtain optical emission corresponding to interband transitions of the LNO/LSMO/STO heterostructures on a near-nanometer scale. By varying incident electron beam energy E_B, one can selectively probe the bulk of each layer as well as the interfaces between them. Figure 1 shows the Monte Carlo simulation for the heterostructure pictured in the inset. Both U₀ and R_B increase with increasing E_B. Discontinuities occur at 30 and 60 nm, corresponding to changes in material density ρ and electron stopping power and are more pronounced at 60 nm due to the larger difference in ρ and atomic number.¹¹ At E_B = 1 keV, excitation occurs in just the top 30 nm LNO layer. E_B = 2 keV excitation occurs in both the LNO and LSMO layers, whereas E_B = 3 keV extends from the free surface into the STO substrate. Figure 1 also shows that, although each distribution has a pronounced peak, there is nevertheless a range of excitation depths on a scale of tens of nm or less. The differential technique we introduce here minimizes DRCLS contributions from depths below U₀. For spectra obtained with excitation at E_B, contributions from these lower depths can be removed by subtracting a normalized spectrum acquired at a lower beam energy. Figure 2 illustrates the DDRCLS method of normalizing the lower energy spectrum within a uniform material. For the same incident beam current I_B, a factor α normalizes the peak of the lower energy (here, 1 keV) Monte Carlo e-h creation rate-versus-depth distribution to the amplitude of the higher energy (here, 2 keV) distribution at the same depth. This subtraction of a lower energy spectrum from a higher energy spectrum results in a more localized Monte Carlo plot and minimal excitation for depths below U₀. Even greater depth resolution can be obtained with smaller energy increments. Note, however, that since the shape of each depth profile varies with E_B, this subtraction method is only approximate. The closer the two E_B are, the more similar the shapes of these profiles are and the greater the resultant depth resolution.

Figure 3(a) shows the CL spectra of an STO wafer obtained from CrysTech Gmbh. There are several peaks both above and below the 3.2 eV (indirect) band gap. The features below 3.2 eV correspond to transitions involving native point defects that have been identified previously. These include the 1.57 eV emission due to Ti³⁺ states^21,22 and the 2.9–3.0 eV peak due to oxygen vacancy-related (V_O-R) defects commonly seen in STO substrates.²³ The broadened peaks are due to different spatial configurations of defects, e.g., interstitial sites and clustering, within the crystal lattice. Features that appear at energies above the indirect band gap of 3.2 eV (Ref. 24) correspond to transitions between orbital-derived energy transitions above the Fermi level and the highest filled states of the valence band as well as the direct Γ-Γ band gap transition at 3.6 eV.²⁵ These transitions are observable because the 5.0 kV incident electron beam has sufficient energy to create e-h pairs not only within the highest occupied valence and lowest unoccupied conduction bands but also within energy levels at much higher empty band levels. Peak widths may correspond to actual orbital broadening due to crystal field splitting and lattice distortions. The peak overlaps may contribute to the large apparent background underlying the partially resolved features. The 0.5 keV differential energies minimize residual artifacts of the subtraction technique due to differences in e-h pair creation rate versus depth profiles.

Figure 3(b) shows the DRCLS spectra of LNO from a “bulk” 60 nm thick LNO/STO sample. Unlike SrTiO₃, LaNiO₃ is metallic so that it has no band gap. The transitions that appear are transitions between orbital-derived energy bands above and at the Fermi level. The transitions were compared with those of previous spectra obtained for these samples. The 2.98 and 3.65 eV peaks that appear can be attributed to a transition between the O_2p and the Ni 3d-derived e_g band, while the 4.13 and 4.69 eV peaks can be attributed to transitions between the La 4f state and Ni 3d t_2g derived bands.²⁶ 

Figure 3(c) shows the DRCLS of synthesized La_2/3Sr_1/3MnO₃ powder, confirmed by X-ray diffraction. As with LNO, LSMO is also metallic²⁷ so that the observed transitions are also those from orbital-derived energy bands above to levels near the Fermi level. The observed 2.26, 2.92, 3.57, 4.13, and 4.64 eV features are attributed to transitions occurring among the Mn 3d e_g and t_2g orbitals.

We used our DDRCLS technique to examine the interfaces of an LNO/LSMO/STO heterostructure. Figure 4(a) shows the interface between the LNO and LSMO films. The 1.0 keV spectrum for the 30 nm LNO layer shows good peak energy correspondence with the 1.5 keV spectrum for the 60 nm LNO on STO in Fig. 3(b). Peak height differences may be due to small differences in lattice distortion that affects orbital symmetries and corresponding transition probabilities. Likewise, the 1.5–1.0 keV differential spectrum of LSMO in Figs. 4(a) and 4(b) show good correspondence with Fig. 3(c) LSMO bulk spectrum. Figure 4(a) exhibits an abrupt transition with no additional peaks that appear only at the interface. In contrast, Fig. 4(b) shows the LSMO/STO interface where two new peaks at 4.01 and 4.50 eV arise. Based on the LSMO and STO peak origins already discussed, we tentatively attribute these features to distortions in the Mn 3d orbitals in the LSMO.

The spectra presented in Fig. 3 show that cathodoluminescence spectroscopy can be a useful tool to measure electronic energy level transitions in ultrathin complex oxides. These spectra show a range of peak features at energies that are well above those generally available with laser excitation. These features can be identified with orbital-derived energy levels derived theoretically. We have observed analogous optical features in STO bulk wafers that are in close agreement²⁸ with theoretical and experimental measurements reported elsewhere.²⁷ The DDRCLS results in Fig. 4 show electronic features that are (i) characteristic of each layer in the multilayer heterostructure as well as (ii) new features that appear at specific interfaces, here the LSMO/STO junction, that correspond to local changes in lattice structure and energy levels. Such changes can alter the energy band lineups across the heterojunction, altering charge transfer between adjoining layers. The abrupt transition between the1.0 keV LNO spectrum and the 1.5 keV interface spectrum in Fig. 4(a) implies that the LNO/LSMO heterointerface grown under the MBE conditions used for LNO/LSMO superlattices is atomically abrupt. In contrast, the spectrum measured by the difference between the interface 3.0 and 2.5 keV LSMO spectra suggest distortions in the LSMO bonding to STO locally. Indeed, while scanning transmission electron microscope images of LNO/LSMO superlattices show atomically abrupt interfaces, slight distortions are sometime noted at the LSMO junction with the STO substrate.^20,29

DDRCLS studies of LNO/LSMO/STO interfaces illustrate its ability to detect optical transitions between orbital-derived energy levels with filled states near the Fermi level of ultrathin complex oxides and to detect changes in the electronic structure at their interfaces on a near-nanometer scale. Further, DDRCLS provides a unique capability to measure electronic features at buried interfaces of ultrathin films. As such, these nondestructive optical measurements can reveal changes in energy levels within nanometers of interfaces, band alignments across interfaces, and the possible effect on those energy levels of local defects.

This work supported by NSF MRSEC Grant No. DMR-1420451 (Charles Ying) and NSF Grant No. DMR-1305193 (Charles Ying and Haiyan Wang). Work at Argonne National Laboratory, including the use of the Center for Nanoscale Materials and Advanced Photon Source, was supported by the U.S. Department of Energy, Office of Basic Energy Sciences under Contract No. DE-AC02-06CH11357. J.D.H. and A.B. acknowledge support from Department of Energy, Office of Basic Energy Science, Materials Science Division.

The Monte Carlo simulation was performed with casino v. 2.48 and its default values. The Z Max profile was used as it details where the electron ends in the sample after creating an e-h pair. We performed calculations with a spot size of 300 μm and simulated 1 000 000 electrons with a sample tilt of 45°. The position of each incident electron X₀ is given by

for both the X and Y positions incident on the sample. R_x is equal to a random number between 0 and 1 with a uniform distribution and d is the diameter of the incident beam on the sample. The distance between the collisions is calculated by

where C_i is the weight fraction of element i and A_i is the atomic weight of element i. ρ is the local density of the material and N₀ is Avogadro's number. The cross section is obtained from a precalculated and tabulated value. The program ignores inelastic scattering and groups all the electron energy loss into a continuous energy loss function. This leads to the following equations for energy loss:

where Z_j and J_j are atomic number and mean ionization potential of element j, respectively. k_j is a variable dependent on Z_j. These steps are repeated until the energy of the electron is below 50 eV or the electron escapes the surface of the sample.

The maximum depth of the electrons that are backscattered from the sample is also included in the simulation by taking the backscattering coefficient into account. By doing this, we can combine the Z Max and Z Max Backscattered profiles using the following formula:

where C_B is the backscattering coefficient that is found in casino for each simulation. This procedure results in the distributions presented in Figs. 1 and 2.