Hard x-ray photoemission spectroscopy of ${\mathrm{LaVO}}_3/{\mathrm{SrTiO}}_3$: Band alignment and electronic reconstruction

A heterostructure consisting of the Mott insulator ${\mathrm{LaVO}}_3$ and the band insulator ${\mathrm{SrTiO}}_3$ is considered a promising candidate for future photovoltaic applications. Not only does the (direct) excitation gap of ${\mathrm{LaVO}}_3$ match well the solar spectrum, but its correlated nature and predicted built-in potential, owing to the nonpolar/polar interface when integrated with ${\mathrm{SrTiO}}_3$, also offer remarkable advantages over conventional solar cells. However, experimental data beyond the observation of a thickness-dependent metal-insulator transition are scarce and a profound, microscopic understanding of the electronic properties is still lacking. By means of soft and hard x-ray photoemission spectroscopy as well as resistivity and Hall effect measurements we study the electrical properties, band bending, and band alignment of ${\mathrm{LaVO}}_3/{\mathrm{SrTiO}}_3$ heterostructures. We find a critical ${\mathrm{LaVO}}_3$ thickness of five unit cells, confinement of the conducting electrons to exclusively Ti $3d$ states at the interface, and a potential gradient in the film. From these findings we conclude on electronic reconstruction as the driving mechanism for the formation of the metallic interface in ${\mathrm{LaVO}}_3/{\mathrm{SrTiO}}_3$.

Figure 1

In the electronic reconstruction model the polar discontinuity at the interface of the LVO/STO heterostructure leads to a potential buildup in LVO. It can be compensated by transferring electrons from the surface to the interface. In LVO/STO, the $3d$ shells of both Ti as well as V could host the electrons.

Figure 2

(a) Monitoring of film growth by RHEED. Regular intensity oscillations of the specular reflex indicate layer-by-layer growth. The RHEED patterns confirm that the substrate and the film surface after the growth of LVO are atomically smooth. (b) X-ray diffractogram of a phase pure 50-uc LVO film on STO. (c) AFM image of a 10-uc LVO/STO sample along with a height profile signaling the good quality of the sample surface. (d) Scanning transmission electron microscopy of a 30-uc LVO/STO sample. Top: HAADF-STEM image showing the interface structure. Bottom: Chemical maps of the same region for substrate and film elements from STEM-EELS. The images confirm a chemically sharp interface and a good structural film quality.

Figure 3

(a) Sheet conductance of the LVO/STO heterostructure at room temperature for varying film thickness. Samples with LVO thicknesses above the critical value of five unit cells show metallic behavior. (b) Sheet carrier concentrations of metallic LVO/STO as well as LAO/STO samples and of the doped Mott insulator LTO. The behavior of the LVO films resembles that of postoxidized LAO/STO. (c) Mobilities of metallic LVO/STO as well as LAO/STO samples.

Figure 4

(a) ResPES at the Ti $L$ edge of a 5-uc LVO film on STO. On resonance, there is spectral weight at the Fermi energy, indicating metallic Ti $3d$ charge carriers (marked by the arrow). (b) ResPES at the V $L$ edge of the same sample. On resonance, the lower Hubbard band (marked as V $3d$ LHB) below the Fermi energy strongly resonates, but there is no spectral weight at the Fermi energy to be seen, indicating the absence of metallic V $3d$ charge carriers.

Figure 5

V $2p_{3/2}$ photoemission spectra of a 6-uc LVO/STO sample for different electron emission angles. The spectra measured with higher angles off normal emission are more surface sensitive. The fact that there is no significant angle dependence suggests a homogeneous distribution of vanadium with a uniform valence over the whole film and accordingly the absence of electron doping into V $3d$ states.

Figure 6

Top: Angle-integrated La $4d$ spectra for three different LVO/STO samples with LVO film thicknesses of 3, 6, and 9 uc integrated over an emission angle $\theta$ of ${(35\pm 28)}^{\circ }$. One can see an energy shift towards lower binding energies with increasing film thickness. Middle: Comparison of the La $4d$ spectrum of the film with a thickness of 6 uc with a reference spectrum of a 30-uc-thick film. The former is broader, consistent with a potential gradient in the film. Bottom: Comparison of the spectrum of the 6-uc LVO/STO sample with model spectra for three different potential gradients. The model spectrum based on a uniform gradient of 0.1 eV/uc best matches the measured spectrum. Also shown is the decomposition of the model spectrum into its single components, corresponding to the six layers. Each single spectrum is exponentially damped according to the depth of its corresponding layer with its energy shift being determined by the gradient.

Figure 7

(a) Top: La $4d$ and V $2p_{3/2}$ spectra for three different LVO/STO samples with LVO film thicknesses of 3, 6, and 9 uc along with the corresponding fit curves. For both core levels, one can see an energy shift towards lower binding energies with increasing film thickness. The fits match the measured spectra fairly well. (a) Bottom: Decomposition of the La $4d$ and V $2p$ fit curves into their single spectra for all three thicknesses. Each single spectrum is exponentially damped according to the depth of its corresponding layer while the energy position is determined by the fitting routine. (b) The potential profiles for each film as deduced from the fitting of the La $4d$ and V $2p$ spectra in (a). The energy shift of the first layer of the 6-uc sample at the interface (layer 1) is set to zero, with a positive shift indicating a shift to a lower binding energy.

Figure 8

(a) Ti $2p_{3/2}$ and (b) Sr $3d$ spectra of a 6-uc LVO/STO sample for different electron emission angles along with a Ti $2p_{3/2}$ reference spectrum of a Nb-doped STO substrate. Beside additional ${\mathrm{Ti}}^{3+}$ spectral weight at lower binding energies (black arrow), both core-level lines show an asymmetry towards higher binding energies as well (red arrows).

Figure 9

Visualization of the interfacial band bending exemplified by a Sr $3d$ spectrum. The photoemission signal of each SrO layer at depth $z$ is assumed to be a symmetric Voigt peak shifted in energy by $E_{\text{bb}}\left(z\right)$. The contributions from deeper layers without an energy shift are treated as a single bulk peak. By assuming a quadratic potential within the bending zone, we can deduce the energetic depth of the band bending $\mathrm{\Delta }{E}_{\text{bb}}$ and its spatial extension $d$.

Figure 10

(a) Exemplary fits for the Ti $2p_{3/2}$ photoemission spectra. The asymmetric line shape can be modeled by several energetically shifted Voigt peaks, representing the individual layers in the bending zone, in addition to the ${\mathrm{Ti}}^{4+}$ bulk peak. Beside the dominant ${\mathrm{Ti}}^{4+}$ peak, there is an additional ${\mathrm{Ti}}^{3+}$ contribution. (b) Exemplary fits for the Sr $3d$ photoemission spectra. For each layer in the bending zone an energetically shifted spectrum is taken into account in addition to a single peak, representing the bulk contribution.

Figure 11

(a) Valence band spectrum of a 6-uc LVO/STO heterostructure and its decomposition into the individual LVO and STO contributions. The valence band onsets are defined by linear extrapolations of the leading edges. (b) Layer-resolved decomposition of the LVO contribution from (a). The individual spectra are shifted in energy according to the potential profile.

Figure 12

Band diagram of the 6-uc LVO/STO heterostructure as deduced from core-level and valence band photoemission, with the value of the optical gap—as defined by the energy difference between the onsets of lower (LHB) and upper Hubbard bands (UHB)—taken from the literature. In the substrate, the bands are bending downwards towards the interface while there is an increasing potential slope in the film. The Ti $3d$ bands, crossing the Fermi level, provide a pocket for electrons to form a two-dimensional electron system (2DES) at the interface. The LHB of the LVO film does not quite reach the Fermi level at the film surface. The bulk valence band offset is zero, since the valence band states of LVO and bulk STO, denoted by VB (“O $2p$”) [56], are aligned.