Effect of SrTiO₃ oxygen vacancies on the conductivity of LaTiO₃/SrTiO₃ heterostructures

Metal oxides are finding many applications in electronic devices as gate dielectrics,¹ thin film semiconductors,² electrochromics,³ magnetic materials, and superconductors,⁴ allowing enhanced performance of conventional devices as well as enabling novel devices that exploit the wide range of functionality. Perovskite oxides, owing to their primarily cubic or pseudocubic structure, provide a versatile class of metal oxide that can be easily integrated as building blocks for next-generation devices. In addition to being closely lattice-matched, several perovskites have been integrated seamlessly with conventional semiconductors by both physical and chemical methods.^5–8 Furthermore, the large array of properties available to these materials can be tuned by the physics inherent to interfaces between these oxides.^9,10 The architecture of transistor, memory, and other microelectronic devices creates many interfaces so that the possibility to use perovskites for both their bulk properties and tunable interfacial phenomena allows for immense versatility in future electronic devices.

A conductive two-dimensional electron gas (2DEG) that forms for specific oxide-oxide combinations is one such interfacial phenomenon that has attracted much attention.^11–18 The prototypical system for 2DEGs in perovskite oxides is the interface between LaAlO₃ and SrTiO₃ (LAO and STO, respectively), first discovered by Ohtomo and Hwang.¹¹ This system has a unique interface in the (001) direction, where the polar layers of LAO (LaO⁺ and AlO₂⁻) create a polar discontinuity with the nonpolar layers of STO (SrO and TiO₂), which forces an electronic reconstruction, and therefore 2DEG formation, to alleviate the incipient diverging potential.^{11,12,16,19–24} This mechanism offers the ability to engineer single interfaces or superlattices with relatively high mobility carriers. Since the discovery of the LAO/STO system, several other mechanisms that promote interfacial conductivity in oxide systems have emerged, including oxygen vacancies^{17,18,25–35} and cation interdiffusion.^33,36 In fact, combinations of several mechanisms have been proposed as contributing to the overall interface behavior.^14,37

The LaTiO₃ (LTO)/SrTiO₃ interface is one system where the origin of the 2DEG is ambiguous. LaTiO₃ (LTO) is polar in the pseudo-cubic (001) direction and, similar to LAO/STO, it is expected to create a polar discontinuity when epitaxially integrated with non-polar STO.^38,39 Several mechanisms can contribute to conductivity in STO and LTO and at their interfaces. Although stoichiometric LTO is a Mott insulator³⁶ at low temperature, a slight excess of oxygen can dope the material and induce conductivity.^40–42 Conversely, oxygen vacancies in STO are a well-studied phenomenon that lead to bulk or interfacial conductivity.^{17,18,30,32,43,44} Furthermore, a small amount of cation interdiffusion essentially creates La-doped STO, a well-known conductive variant of STO.^33,36 Finally, LTO has been shown to undergo a metallic transition under sufficient strain, and the lattice mismatch between LTO (a = 3.97 Å) and STO (a = 3.905 Å) is enough to cause the insulator-to-metal transition for LTO thin films, delta-doped STO, or LTO/STO superlattices.^45–51 Strain engineering LTO/STO heterostructures may prevent this particular mechanism from occurring. Specifically, the lattice mismatch between pseudocubic LTO and the interatomic distance of 2 × 1 Ge(001) surface dimers along <110> (3.992 Å) is only 0.5% and a thin, strained STO template layer may preserve this small lattice mismatch while allowing epitaxial growth of the heterostructure.

In this work, we report the strong influence of growth conditions on the conductivity of the LTO/STO and LTO/STO/Ge systems, in particular elucidating the role of oxygen vacancies that are created in STO and the conductive layer formed through these vacancies. We employ a combination of X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), and transmission electron microscopy (TEM) for LTO/STO and LTO on STO-templated Ge(001) to determine the significant presence of oxygen vacancies in the LTO/STO heterostructures. Through electrical measurements, we show that the conductivity of these systems is strongly tied to the growth conditions, which heavily influence the formation of oxygen vacancies in the STO and the incorporation of oxygen in the LTO. Changing the growth conditions offers the ability to tailor the conductivity of the underlying substrate or layer.

TiO₂-terminated STO(001) 5 mm × 5 mm × 0.5 mm single-crystal substrates (CrysTec GmbH) underwent a degreasing procedure of ultrasonication in acetone, isopropanol, and deionized water for 5 min each. The substrates were then exposed to UV/ozone for 15 min and subsequently loaded into the deposition system. The custom-built system is described in detail elsewhere.^52,53 Immediately prior to growth, substrates were annealed at 750 °C for 30 min under ultra-high vacuum (UHV). Film growth was carried out in a DCA 600 molecular beam epitaxy (MBE) chamber in which La and Ti were thermally evaporated from effusion cells. Metal fluxes were measured with a quartz crystal microbalance inserted just below the substrate, and they were calibrated to achieve a 1:1 atomic ratio as determined by X-ray photoelectron spectroscopy (XPS). LTO films were grown by co-depositing La and Ti onto the STO substrates heated to 700 °C, unless specified otherwise, under partial pressures of molecular oxygen ranging from less than 10⁻¹⁰ Torr to 10⁻⁷ Torr.

The Ge(001) substrates (MTI Corporation, Sb-doped, 0.1 Ω cm) were prepared by dicing the 1.5-in. wafer into 10 mm × 10 mm pieces. Ge substrates underwent the same degreasing procedure and UV/ozone exposure as the STO substrates. After loading into the deposition system, the Ge native oxide was thermally desorbed by annealing at 700 °C for 1 h under UHV. Upon cooling to 200 °C, the Ge surface showed a 2 × 1 reconstruction as verified by in situ reflection high-energy electron diffraction (RHEED). This treatment had been shown previously to yield high quality Ge surfaces for subsequent deposition.^6,54–56 STO templates were grown on the Ge substrates via MBE by an established process of depositing 5 unit cells (u.c.) of amorphous STO, using alternating Sr and Ti deposition at 200 °C under molecular oxygen at a pressure of 5 × 10⁻⁷ Torr.⁵⁷ The substrates were heated to 750 °C, during which the STO crystallized at ∼550 °C. Subsequent STO growth was carried out at 750 °C using 5 × 10⁻⁶ Torr of molecular oxygen. LTO thin films were then deposited on the STO template layers at 550 °C under different partial pressures of molecular oxygen. The difference in LTO growth temperature between STO substrates and Ge substrates was necessitated by the geometrical difference in sample holders (5 mm × 5 mm versus 10 mm × 10 mm) and the emissivity differences of the substrates, which affected the steady-state temperatures achieved by the substrates. The temperatures reported here are the set-point temperatures of the silicon carbide heater, which is in close proximity to the sample holder.

Film morphology was monitored in real time during growth using RHEED operated at 21 keV and at an incident angle of 3° (Staib Instruments). After growth, films were transferred in situ to an analysis chamber to determine the elemental composition using XPS (VG Scienta R3000 analyzer with monochromatic Al Kα radiation and photon energy of 1486.6 eV). XPS peak fitting and deconvolution was performed using CasaXPS software (version 2.3.16 PR 1.6). Transmission electron microscopy (TEM), X-ray diffraction (XRD), and X-ray reflectivity (XRR) were employed to characterize the film crystallinity and thickness. The TEM studies, as well as electron-energy-loss spectroscopy (EELS) analysis, used a probe-corrected JEOL ARM200F scanning transmission electron microscope operated at 200 keV. High-angle annular-dark-field (HAADF) images were recorded with inner and outer collection angles of 90 and 150 mrad, respectively. XRD and XRR were performed with a Rigaku Ultima IV diffractometer using Cu Kα radiation (λ = 1.5406 Å) at 40 kV and 44 mA. XRD studies of LTO films on STO substrates were performed with a Ge(220) two-bounce monochromator in order to distinguish the substrate and film peaks. XRD data fitting to determine lattice parameters was performed using 1D or 2D Gaussian functions, depending on the scan type, in IGOR Pro software (version 6.3.7.2).

Electrical measurements were conducted in a physical property measurement system (PPMS) by Quantum Design Inc. capable of applying a magnetic field up to ±9 T over a temperature range of 1.7 K–350 K. In order to electrically contact the LTO/STO interface, four indium contacts were placed on scribed corners of each sample in van der Pauw geometry except for LTO/STO/Ge samples, on which the indium contacts were directly placed without being scribed. Two lock-in amplifiers (SRS SR830 and NF LI5640) and one current preamplifier (DL Instruments, LLC) were used to measure current I and four-probe voltage, V_xx, or Hall voltage, V_xy, with the standard four-wire electrical transport measurement method. For all measurements, a current less than 10 nA was applied to the sample at a frequency of 7 Hz. Four-probe resistance, R_xx, is defined as V_xx/I, and sheet resistance, ρ, is equal to πR_xx/ln 2 according to the van der Pauw method. R_xx, Hall mobility, and sheet carrier density were measured as a function of temperature.

Representative RHEED patterns of a bare TiO₂-terminated STO(001) substrate and a 12-nm-thick LTO film grown on an TiO₂-terminated STO(001) substrate are shown in Figs. 1(a) and 1(b). RHEED monitoring during LTO growth showed that the films were crystalline when deposited at 700 °C. The patterns depict the [110] azimuth of the perovskites, and the sharp pattern, along with the clear Kikuchi lines and second-order spots, indicates a highly crystalline LTO film with low surface roughness. The faint pattern visible between the [110] diffraction spots corresponds to the expected √2 × √2 R45°-reconstructed surface pattern that results from LTO relaxation to its bulk orthorhombic structure (space group Pbnm) near the film surface.⁵⁸ Figure 1(c) shows the out-of-plane XRD pattern of an 8-nm-thick LTO film on STO, which shows only pseudo-cubic (00l) peaks and confirms that the films are single-phase. The LTO(002) peak position corresponds to an out-of-plane pseudo-cubic lattice parameter of 4.018 Å. Bulk pseudo-cubic LTO (a = 3.97 Å) has a 1.6% lattice mismatch with STO (a = 3.905 Å) and causes biaxial in-plane compressive strain in the LTO that accounts for the observed elongation of the out-of-plane lattice parameter.⁵⁹ In order to further confirm the LTO crystal structure and epitaxy with the STO substrate, cross-sectional TEM was employed. Figure 2 is a HAADF STEM image showing a different LTO/STO heterostructure. The strong contrast difference between the atomic columns of La and Sr provides clear evidence of an abrupt interface and confirms the highly epitaxial nature of the LTO film. The LTO film appears to be highly ordered up to ∼16-20 u.c. from the interface, beyond which some regions appear to become somewhat disordered but regain crystallinity as the growth continues.

To eliminate potential strain-induced metallic behavior and to limit the amount of oxygen available from the substrate, thin LTO films were also grown on thin STO templates strained to Ge(001) substrates. Films were crystalline as-deposited at 550 °C with 10⁻⁷ Torr of molecular oxygen. The Ge-Ge separation along the [110] direction is 3.992 Å which provides sufficient (2.1%) tensile strain to the STO template to prevent strain-induced metallicity in the LTO film. Epitaxial registry in this LTO/STO/Ge heterostructure is confirmed by HAADF STEM imaging (Fig. 3). Interestingly, the LTO film appears to have a disordered region similar to that in the LTO/STO heterostructure shown in Fig. 2, albeit at a much shorter distance from the interface. While the exact nature and reason for this structural rearrangement are beyond the scope of this work, this rearrangement is likely caused by slight over-oxidation in the film that disrupts local crystallinity.

Reciprocal space maps (RSMs) were taken for LTO/STO and LTO/STO/Ge heterostructures to characterize the strain state of the thin films. Figure 4(a) shows an RSM taken about the LTO/STO(103) diffraction peaks for an 8-nm-thick LTO film on an STO substrate. The LTO(103) peak is clearly aligned in-plane with substrate peak, indicating that the film is strained to the substrate. The LTO(103) peak center corresponds to in-plane and out-of-plane lattice parameters of 3.906 Å and 4.024 Å, respectively. Figure 4(b) shows an RSM taken about the LTO/STO(103) diffraction peak for the sample shown in Fig. 3. The STO peak location confirms that it is strained to Ge with lattice parameters of a = 3.981 Å and c = 3.883 Å. The LTO peak position corresponds to lattice parameters of a = 3.973 Å and c = 3.991 Å, which are close to the pseudocubic lattice constant of 3.97 Å for bulk, unstrained LTO.⁵⁹ 

XPS was conducted in situ, without exposure of the films to ambient atmosphere, to confirm the film stoichiometry as well as to determine the possible existence of oxygen vacancies. The Ti 2p core-level electrons of optimally oxidized LTO (i.e., fully Ti³⁺) exhibit final-state effects that are manifested in X-ray photoelectron (XP) spectra as a two-component split of Ti⁴⁺ and Ti³⁺ features.⁶⁰ The Ti 2p X-ray photoelectron (XP) spectrum shown in Fig. 5(a) for a representative LTO film grown under background oxygen partial pressure (<10⁻¹⁰ Torr) reflects this final state that, in combination with La 3d and O 2p XP spectra (not shown), confirms the desired 1:1:3 La:Ti:O film stoichiometry. However, even minor introduction of oxygen during LTO growth served to over-oxidize the films when grown on STO. Figure 5(b) illustrates the effects of increasing oxygen partial pressure during LTO growth. As the oxygen pressure is increased, the ratio of Ti⁴⁺:Ti³⁺ signals changes until the film consists of nearly all Ti⁴⁺, corresponding to a film stoichiometry of LaTiO_3.5, for oxygen partial pressure above 1 × 10⁻⁷ Torr. Films grown under increased partial pressures of oxygen were amorphous based on RHEED patterns diminishing completely after ∼4 min of deposition. Amorphous films were rather insulating and electrical transport data could not be gathered from them.

Figures 5(a) and 5(b) suggest that the oxygen made available by scavenging from the STO substrate provides an adequate source for the growing LTO films. The introduction of oxygen gas to the growth chamber, even at low partial pressure, serves to create an over-oxidizing environment for the LTO films. Oxygen vacancies have been shown to form at the interfaces of STO in other metal oxide/STO systems, involving much less reducing environments than this study.^17,18,61 Furthermore, oxygen diffusion in perovskites, and STO in particular, has been repeatedly shown to occur through a vacancy mechanism with relatively high diffusion coefficients.^62–66 Thus, the growing LTO film likely scavenges oxygen from the STO surface region, while oxygen within the STO substrate diffuses toward the interface to equilibrate the vacancy concentration.⁶⁷ 

More credence is given to this explanation by a comparison of the oxygen availability between the O₂ gas in the growth chamber and the oxygen within the STO substrate. Assuming an average gas molecule temperature of 300 K and using a surface density of STO and LTO atoms of ∼10¹³ cm² [for (100) planes], we predict the time required for one monolayer equivalent of molecular oxygen to reach the surface of the STO/LTO sample.⁶⁸ These values are listed in Table I for several different partial pressures of molecular oxygen. Assuming one-dimensional transport of oxygen through the STO substrate and using a reported diffusion coefficient of ∼10⁻⁵ cm²/s for oxygen isotope tracers in STO at 750-850 °C, the average lengths traveled by an oxygen atom in the STO substrate for the previously calculated times are also provided in Table I.⁶² For reference, the La and Ti deposition rates were calibrated to provide ∼1 u.c. LTO per minute. Partial pressures of 10⁻¹⁰ Torr or lower cannot provide enough oxygen to the surface of the growing film. In contrast, an oxygen atom within the STO can travel many thousands of nm within seconds and supports the conclusion that the substrate can readily supply oxygen to the growing LTO film. At 10⁻⁹ Torr, monolayer coverage from oxygen gas reaches the same time scale as the deposition rate and, in combination from oxygen diffusion from the substrate, is likely the cause of the beginning of over-oxidation of the LTO film shown in Fig. 5(b). At 10⁻⁸ Torr or higher, oxygen gas supplies enough oxygen to the growing film to over-oxidize it even further.

Evidence for the presence of Ti³⁺ and possible oxygen vacancy segregation in the vicinity of the LTO/STO interface was provided using EELS, in particular through energy-loss near-edge structure (ELNES) analysis, which enables mapping of the Ti oxidation state across the interface.⁶⁹ The ELNES line profile in Figs. 6(a)–6(c) shows the results of applying signal-from-background separation⁷⁰ and hyper-spectral unmixing analysis of ELNES to an as grown (<10⁻¹⁰ Torr molecular oxygen) LTO/STO heterostructure.⁷¹ A description of the hyper-spectral unmixing method is found elsewhere.⁷¹ Figures 6(a) and 6(b) are a line-profile and a two-dimensional map based on the ELNES analysis showing the intensity of the three different species represented in the Ti-L edge EELS signal. The ELNES spectra of the three different species are shown in Fig. 6(c), using the same color as in (a) and (b). The red spectrum showing well-separated peaks is Ti⁴⁺ in STO. The green spectrum showing two broad peaks is Ti³⁺ either in STO or LTO. The blue spectrum, which shows slight splitting of the e_g and t_2g peak, indicates the over-oxidized Ti in LTO. The exact valence is not determined from the fine structure due to other factors such as local strain and lattice distortions as well as the parameters of the microscope itself that also introduce small changes. For this reason, we use Ti^3+δ to indicate over-oxidized Ti in LTO (shown in dark blue).

In Fig. 6(a), low contrast in the corresponding HAADF image (not shown) in the LTO far from the interface, in combination with the slight splitting of the e_g and t_2g peaks of the dark blue spectrum in the Ti-L edge [Fig. 6(c)], indicates lower crystallinity caused by over-oxidation of the LTO, especially in the right part of the image (upper part of the film). The difference in the de-localization effect observed in STEM-EELS should be less than 0.1 nm for Ti-L and La-M edges.⁷⁴ However, the Ti³⁺ (green) peak of the Ti-L edge also extends into the STO substrate about 0.2 nm deeper than the elemental signal from the La M-edge (orange), indicating the presence of a very thin layer of high Ti³⁺ concentration on the STO side of the interface, which would also contribute to the conductivity. This Ti³⁺ feature may originate from charge modulation doping⁷² or oxygen vacancies. The depth of Ti³⁺ layer in charge modulation doping is small in both experimental and theoretically calculated results.^38,72,73 Our result shows that the interfacial Ti³⁺ region only ∼1 nm in thickness is consistent with the reported values. It corresponds to a sheet carrier density of ∼10¹⁴/cm² at most, which is much smaller than the electrical measurements described later.¹¹ Hence, carriers deep in the bulk of the STO substrate, possibly introduced by oxygen vacancies, must also contribute to the conductivity. However, the measurement of oxygen vacancies deeper in the bulk STO is beyond the detection limit of the ELNES mapping analysis and we estimate that the concentration of oxygen vacancies in the substrate is less than 1%, which is discussed later with the results from the electrical measurements.

Although STO substrates can act as nearly infinite sources of oxygen, a thin layer (<10 nm) of STO deposited on Ge can only provide a finite amount of oxygen from which subsequent LTO growth can draw. Figure 7(a) compares the Ti 2p XP spectra from LTO growth on STO templates on Ge(001) with and without oxygen gas flowing through the chamber. At 10⁻⁷ Torr of oxygen, the LTO film is slightly over-oxidized (LaTiO_3.2), which is evident from the pronounced Ti⁴⁺ feature at ∼459 eV. However, LTO growth at chamber background oxygen levels (<10⁻¹⁰ Torr) results in a combination of oxidized Ti species and a zero-valent Ti signal emerging at ∼454 eV. The Ti 2p spectra of the STO layer prior to La/Ti deposition are shown for a reference in Fig. 7(a). La 3d spectra (not shown) also exhibit a mixture of oxide and metal peaks. Figure 7(b) compares the corresponding Ge 3d XP spectra. The STO layer is thin enough that the Ge peak is still visible at the expected binding energy shift observed in Ge-Sr Zintl layers.⁵⁶ In the case of La/Ti deposition without oxygen, the Ge signal remains but is shifted toward zero-valent Ge (or possibly La or Ti germanides). Despite the presence of the Ge signal, the Sr 3d signal vanishes under these conditions (not shown). These observations suggest extensive reduction of the STO and subsequent desorption of SrO or Sr species, both of which are volatile at the growth temperature.⁷⁵ 

The strong thickness-dependence of the electrical transport properties of LTO/STO heterostructures is illustrated in Fig. 8 for samples grown on STO substrates at molecular oxygen partial pressures <10⁻¹⁰ Torr. Several LTO films were capped with 2.5-nm-thick Al₂O₃ films grown by atomic layer deposition (ALD) at 175 °C to protect the LTO films from atmospheric oxidation. A description of the ALD-grown alumina layers has been reported elsewhere.¹⁷ This over-layer had little effect on films thicker than 4 u.c. However, a 2-u.c. LTO film was too insulating to measure without a capping layer. This could be explained by the ALD process oxidizing a 2-4-u.c. layer of LTO near the LTO/alumina interface and causing LTO conductivity through the oxygen excess.^40–42 At room temperature, sheet resistances [Fig. 8(a)] decrease by as much as a factor of 10³ between 2-u.c. LTO with 2.5-nm-thick Al₂O₃ and 20-u.c. LTO, whereas resistances at 10 K are up to a factor of 10⁵ lower for the thicker films. The origin of the feature in the uncapped, 4 u.c. LTO film resistivity curve is unknown, although it is likely due to poor or damaged film quality from the relatively low film thickness and exposure to atmosphere, which is supported by the fact that an uncapped, 2 u.c. film could not be measured due to poor electrical transport. Low-temperature Hall mobility increases significantly for LTO thicknesses below 8-u.c. and reaches ∼10⁴ cm²/V s at 10 K for 8, 16, and 20-u.c. LTO while all films are on the order of 1 cm²/V s at room temperature [Fig. 8(b)]. The high values of the Hall mobility are comparable to La- and Nb-doped and reduced STO single-crystals.^76–78 The sheet carrier density is approximately constant with temperature and increases monotonically with film thickness, strongly suggesting bulk conduction in both the STO and LTO layers [Fig. 8(c)]. Given that a 2-u.c. LTO film without a capping layer is insulating (not shown), the minimum carrier density needed to observe conductivity in this system must be ∼10¹³ cm⁻². Again, it should be noted the conductivity of the 2-4 u.c. films could be heavily influenced by the possibility of LTO oxidation that could occur during ALD growth of the alumina capping layer.

The 20-u.c. (8-nm-thick) LTO sample exhibits a sheet carrier density of ∼5 × 10¹⁶ cm⁻², which is greater than can be attributed to any one mechanism in the absence of oxygen vacancies. Even vacancies confined to the STO interface cannot explain this high carrier density, as it would require an impossible ∼250% vacancy of oxygen sites for a 10 u.c. thick layer of STO to generate the same number of carriers. Considering the detection limit of ELNES mapping analysis is about 1%, we estimate that oxygen vacancies reach at least ∼1000 nm into the STO substrate. The high diffusion coefficient of oxygen in STO makes it likely that the vacancies extend much further than this value and that the vacancy density is less than 1%, which is in agreement with the ELNES analysis.⁶² Systems exhibiting polar catastrophe typically have limiting carrier densities on the order of 10¹⁴ cm⁻² per electronic reconstruction.¹⁵ Materials with oxygen vacancies confined to the STO interface have carrier densities on the order of 10¹³-10¹⁴ cm⁻².^17,18 Comparable thicknesses of strained LTO would only account for sheet carrier concentrations on the order of 10¹⁵ cm⁻².⁴⁵ Although the LTO/STO heterostructures likely have a strain-based contribution to the electrical characteristics, the carrier concentration of such a contribution still remains an order of magnitude lower than the observed results. Carrier densities similar to those shown in Fig. 8(c) are observed for systems that generate oxygen vacancies in bulk STO, which supports the suggestion that oxygen vacancies are being generated in the STO substrate because of the LTO deposition and account for the majority of carriers observed in this work.²⁵ 

To further evaluate the impact of oxygen vacancies, a 30-u.c. LTO/STO heterostructure (grown at 850 °C) was annealed in 5 × 10⁻⁷ Torr oxygen at 850 °C for 30 min. Figures 9(a)–9(c) compare the electrical properties of this annealed film with a film grown under identical conditions but without the oxygen annealing step. The sheet resistance is higher for the annealed film across the entire temperature range, whereas the carrier density decreases by a factor of about 2/3, suggesting that some oxygen vacancies are healed. Longer or more oxidizing annealing conditions caused over-oxidation of the LTO film, which itself could induce LTO metallic behavior for particular amounts of oxygen doping.^40–42 To discount the effects of over-oxidation of the LTO films, a simple resistance measurement was performed with a multimeter contacting silver paste on the surface of the LTO films. The non-annealed film had a two-point resistance on the order of 5 × 10⁵ Ω, while both the annealed film and LTO/STO/Ge film stack had resistances approximately one order of magnitude lower. These measurements indicate that the oxidizing conditions for the latter two films may impart minor conductivity to the LTO film and that non-annealed films exhibit conductivity that is dominated by the STO substrate. Although Hall mobility was not measured at temperatures down to 10 K, one may assume approximately constant carrier densities and predict a Hall mobility of ∼10⁴ cm²/V s for the annealed film. Again, films grown on STO substrates and increased partial pressures of oxygen were over-oxidized and amorphous, and they were very insulating in electrical measurements.

The LTO/STO/Ge system exhibited distinctly different electrical properties, most likely because of suppressed oxygen vacancy formation in the thin STO template as well as the elimination of any strain-induced metallic behavior of the LTO film. Recall that higher partial pressure of molecular oxygen was required to achieve growth of crystalline LTO on the STO/Ge heterostructure and likely suppressed the formation of oxygen vacancies. At room temperature, the sheet resistance [Fig. 10(a)] and Hall mobility [Fig. 10(b)] reflect the Sb-doped Ge substrate. The Ge conduction freezes out at low temperature and the sheet resistance increases to ∼100 Ω, on par with the 2-u.c. LTO/STO heterostructure with a 2.5-nm-thick Al₂O₃ capping layer. The Hall mobility drops sharply at about 200 K and then levels off at an approximately constant 5 cm²/V s, which is much lower than for any other films measured in this work, likely due to the extremely thin STO layer relative to the samples grown on 0.5 mm thick STO substrates. The carrier density experiences a shallow maximum around 150 K but remains on the order of 10¹⁵ cm⁻² across the entire temperature range. Albeit higher than the thinnest LTO films on STO substrates, the carrier density of the LTO/STO/Ge stack is roughly an order of magnitude lower than for films of the same thickness grown on STO substrates, and is attributed to the relative absence of oxygen vacancies in STO because of the higher partial pressure of oxygen in the chamber required to grow crystalline LTO.

Although the LTO/STO/Ge heterostructure relieves strain in the LTO layer and provides evidence of oxygen scavenging from the underlying STO layer, the crystalline quality of the LTO layer makes further study of the heterostructure difficult. A minimum thickness of 4 u.c. is needed to observe a polar discontinuity in prototypical perovskite heterostructures.⁷⁹ Figure 3 shows significant contrast variation in the LTO layer after 2 u.c. and suggests that the heterostructure should not be relied upon for the study of electrical transport properties dependent on high quality LTO single crystals. Further work is needed to control and optimize the growth conditions of the LTO layer for future study of this system.

This work demonstrates the growth of crystalline LTO on both STO single-crystal substrates and STO templates on Ge. XRD and TEM analyses confirm single-phase LTO and high quality interfaces between LTO and STO. XPS measurements of Ti 2p core level electrons further indicate stoichiometric LTO films. Comparison of Ti 2p and Ge 3d core levels between LTO/STO and LTO/STO/Ge systems confirms the presence of oxygen vacancies due to La/Ti oxygen scavenging from the STO during the LTO film growth. Electrical measurements show a strong dependence on the LTO film thickness, further supporting the existence of oxygen vacancies. Although other mechanistic pathways may be contributing to the conductivity of these heterostructures, any one or combination of these pathways cannot account for the observed carrier densities. Upon mild oxygen annealing, or for growth on thin STO buffer layers on Ge substrates, electrical behavior is suppressed due to healing or prevention of oxygen vacancies in STO and, in the latter case, also the elimination of potential strain-induced metallic behavior in LTO. This work demonstrates the potential for engineering thin conductive layers of STO through careful control of LTO growth conditions.

This work was supported by the Office of Naval Research (Grant No. N00014-10-10489) and the Air Force Office of Scientific Research (Grant Nos. FA9550-12-10494 and FA9550-12-1-0441). D.J.S., S.L., and H.W. acknowledge the use of facilities in the John M. Cowley Center for High Resolution Electron Microscopy at Arizona State University.