Report

Metallic and Insulating Oxide Interfaces Controlled by Electronic Correlations

See allHide authors and affiliations

Science  18 Feb 2011:
Vol. 331, Issue 6019, pp. 886-889
DOI: 10.1126/science.1198781

Abstract

The formation of two-dimensional electron gases (2DEGs) at complex oxide interfaces is directly influenced by the oxide electronic properties. We investigated how local electron correlations control the 2DEG by inserting a single atomic layer of a rare-earth oxide (RO) [(R is lanthanum (La), praseodymium (Pr), neodymium (Nd), samarium (Sm), or yttrium (Y)] into an epitaxial strontium titanate oxide (SrTiO3) matrix using pulsed-laser deposition with atomic layer control. We find that structures with La, Pr, and Nd ions result in conducting 2DEGs at the inserted layer, whereas the structures with Sm or Y ions are insulating. Our local spectroscopic and theoretical results indicate that the interfacial conductivity is dependent on electronic correlations that decay spatially into the SrTiO3 matrix. Such correlation effects can lead to new functionalities in designed heterostructures.

Advanced deposition techniques enable the growth of epitaxial heterostructures with atomically controlled interfaces such as multilayers (1), superlattices (24), and ultrathin films (5, 6). In these artificial structures, the interfaces play a prominent role in determining the functionalities of the structures and their applications (7). A recent example is the discovery of two-dimensional electron gases (2DEGs) at the interface between complex insulating oxides (8) such as LaAlO3/SrTiO3 (9, 10), LaTiO3/SrTiO3 (2), and LaVO3/SrTiO3 (11) heterostructures, in which the 2DEG is confined near the LaO/TiO2 interface. Magnetic and superconducting ground states of the 2DEG have been identified (1214), and applications to field-effect transistors and tunnel junctions have been demonstrated (1517).

Theoretical work on LaTiO3/SrTiO3 superlattices (18) suggests that for a several-unit-cell-thick LaTiO3 layer, the LaTiO3/SrTiO3 interface region is metallic; however, nonmetallic behavior dominates in the LaTiO3 region away from the interface, resulting from strong electron correlations similar to those found in bulk LaTiO3. In other bulk rare-earth titanates, the effect of electron correlations depends critically on the rare-earth ion (19). We used the unique electronic character of oxide interfaces, and atomic level control of their structure and composition, to deliberately manipulate the 2DEG electronic properties.

We studied the effect of strong electron correlations on an oxide 2DEG by inserting a single atomic layer of RO (R is La, Pr, Nd, Sm, or Y) into an epitaxial SrTiO3 matrix using pulsed-laser deposition with atomic layer control. The RO layer donates electrons to the conduction band of SrTiO3. These electrons remain near the inserted RO layer due to Coulomb attraction. We find that the transport properties of these electrons range from metallic to insulating, depending critically on the rare-earth ion, and that this dependence arises from strong electronic correlations.

We grew epitaxial SrTiO3 heterostructures containing a symmetric TiO2/RO/TiO2 interface (Fig. 1A), resulting in RTiO3-like structure at the interface. Using pulsed-laser deposition, the heterostructures were fabricated by depositing either a RO monolayer or a RTiO3 unit cell on a TiO2-terminated SrTiO3 substrate, followed by deposition of a SrTiO3 overlayer of varying thickness (20). A thick SrTiO3 overlayer approximates a single RO monolayer embedded in an infinite SrTiO3 matrix. Thicknesses of inserted 1-monolayer (ML)–thick RO and 1-unit-cell (uc)–thick RTiO3 layers were accurately controlled by monitoring in situ reflection high-energy electron diffraction (RHEED) intensity oscillations. Typical RHEED oscillations for the growth of a 10-uc SrTiO3/1-ML LaO heterostructure on a SrTiO3 substrate are shown in Fig. 1B. The atomic force microscopy (AFM) image of the surface of a complete heterostructure (Fig. 1C) shows the steps and terraces of the original substrate surface, indicating high-quality growth. Microstructure and electrical properties of both SrTiO3/1-ML RO/SrTiO3 and SrTiO3/1-uc RTiO3/SrTiO3 heterostructures were almost identical (20). Here, we focus on the SrTiO3/RO/SrTiO3 heterostructures.

Fig. 1

(A) Schematic representation of a SrTiO3/1-ML RO/SrTiO3 heterostructure. The atomic structure near the interface is enlarged. The +1 valent RO layer donates electrons to neighboring TiO2 planes, leading to the larger electron density ne near the interface. (B) Typical RHEED oscillations for the growth of 1-ML LaO and 10-uc SrTiO3 layers in sequence on a TiO2-terminated SrTiO3 substrate. (C) AFM image of a 10-uc SrTiO3/1-ML LaO/SrTiO3 heterostructure showing an atomically smooth surface.

We first characterized the dependence of electrical properties on growth conditions, using the LaO-based heterostructure, and established the growth conditions of oxygen pressure (PO2=103mbar) and temperature (Tgrowth = 550°C) as optimal (20). These growth conditions were used to fabricate SrTiO3 heterostructures with single inserted atomic layers of LaO, PrO, NdO, SmO, and YO. Fig. 2A shows the mobile sheet carrier concentration ns for the five different RO layers as a function of the SrTiO3 overlayer thickness. It is seen that LaO-, PrO-, and NdO-based heterostructures become conducting above the critical thickness of SrTiO3 of three or four unit cells. However, SmO- and YO-based heterostructures are insulating, even with a SrTiO3 overlayer thickness of 100 uc. This is summarized in Fig. 2B, which shows the mobile sheet carrier concentration at fixed overlayer thickness as the rare-earth ion progresses from La to Y. The nominal room-temperature concentration of mobile carriers in crossover NdO-based heterostructures decreases dramatically at lower temperatures (fig. S3D), in contrast to the relatively temperature-independent behavior of the conducting LaO-based and PrO-based heterostructures. This trend is analogous to that in bulk RTiO3, where the effects of electron correlations increase as R is varied from La to Y (21). The mobilities of all conducting heterostructures are roughly independent of the rare-earth ion, showing a crossover from temperature-dependent phonon scattering at high temperature to a temperature-independent value at low temperatures.

Fig. 2

Dependence of sheet carrier concentration ns on the R ion in SrTiO3/1-ML RO/SrTiO3 heterostructures and the SrTiO3 overlayer thickness d. Sheet carrier concentration is plotted as a function of (A) the thickness of the SrTiO3 overlayer and (B) the RO doping layer for SrTiO3/1-ML RO/SrTiO3 heterostructures. SmO-based and YO-based heterostructures never become conducting, even with very thick SrTiO3 overlayers.

Our transport measurements are sensitive to mobile carriers near the interface. We also investigated charge transfer from the RO layer to nearby Ti states with electron energy-loss spectroscopy (EELS), sensitive to both mobile and nonmobile carriers (2, 10). For a conducting LaO-based heterostructure, the spatial dependence of EELS spectra of Ti-L2,3 and O-K edges is shown in Fig. 3B. The EELS spectra are spatially separated by 0.28 nm, in a line scan across the interface of a 10-uc SrTiO3/1-ML LaO/SrTiO3 heterostructure (Fig. 3A). Four clear peaks in the Ti L2,3 edge become broader at the interface, with peak separations less pronounced. We attribute this broadening to the presence of a Ti3+ component. Compared with previous reports (2, 10), the relatively small modulation of the EELS signal at the interface may be related to the low ns determined from the Hall effect. Our depth profiling of the Ti3+ to Ti4+ ratios indicates that the carriers are confined to within ~1 nm of the interface (fig. S6), in good agreement with recent theoretical calculations (22).

Fig. 3

STEM and EELS analysis. (A) High-angle annular dark field (HAADF) image of a 10-uc SrTiO3/1-ML LaO film grown on SrTiO3. The rectangular box represents the region of EELS line scans. (B) EELS spectra of T-L2,3 and O-K edges obtained from 2D line scans across the interface shown in (A). The spacing along the line scan between consecutive EELS spectra is 2.8 Å. The spectra at the LaO layer are highlighted by thicker lines. For the spectra for Ti L2 and L3 edges, peak broadening and less pronounced peak splitting at the interface are clearly observed. (C) HAADF images of 10-uc SrTiO3/1-ML LaO/SrTiO3 and 10-uc SrTiO3/1-ML SmO/SrTiO3 heterostructures. Both samples show no obvious defects or dislocations, indicating coherent interfaces. (D) Selected area Ti-L2,3 EELS spectra obtained at the interfaces for 10-uc SrTiO3/1-ML LaO/SrTiO3 and 10-uc SrTiO3/1-ML SmO/SrTiO3 heterostructures. The arrow is a guide for comparison.

Fig. 3, C and D, show scanning transmission electron microscope (STEM) images and selected area Ti-L2,3 EELS spectra, at TiO2 planes adjacent to the interface, for LaO (conducting) and SmO (insulating) heterostructures. For both heterostructures, the STEM images and the Ti-L2,3 spectra at the interface look very similar. In particular, the very similar peak splittings at ~462 eV in the Ti L2 edges suggest that the electron transfer from the RO layer to the neighboring TiO2 planes is the same for both LaO- and SmO-based heterostructures. Our transport measurements indicate that these electrons produce a conducting 2DEG in LaO heterostructures but are not mobile in SmO heterostructures.

TiO6 octahedra rotations in bulk RTiO3 determine the width of the Ti-3d band of t2g symmetry, and hence the electronic properties, through a change in the Mott-Hubbard gap (23). SrTiO3, however, has no TiO6 octahedral rotations at room temperature. We investigated octahedral rotations in our SrTiO3/RO/SrTiO3 heterostructures, with synchrotron x-ray experiments at the Advanced Photon Source. We observed strong superlattice reflections (figs. S7 and S8) resulting from unit-cell doubling TiO6 octahedra rotations, in good agreement with the density functional calculations discussed below. The octahedral rotations are well ordered in the interfacial plane, with typical rocking widths giving an in-plane domain size > 60 nm. The breadths of the half-order peaks in the out-of-plane direction are consistent with octahedral rotations at the RTiO3 layer rapidly decaying into the SrTiO3 matrix. These decaying octahedra rotations lead to a spatial gradient in the electronic structure, influencing the conduction.

In addition, epitaxial strain in the interfacial RTiO3 layer also affects the interface conductivity. LaTiO3, PrTiO3, and NdTiO3 layers at the interface are strained under biaxial compression, but SmTiO3 and YTiO3 layers are under biaxial tension (table S1) (21). Compressive strain has been shown to induce conducting behavior in LaTiO3 thin films (24), attributed to an increased Ti t2g bandwidth and a weakened crystal field. This has been predicted theoretically to reduce the effect of electron correlations and to support metallic behavior (25). The tensile strain in the SmTiO3 and YTiO3 layers embedded in SrTiO3 appears to enforce the effect of strong correlations and favor insulating behavior.

To understand the combined effects of charge transfer, spatially varying octahedral rotations, biaxial strain, and rare-earth electronic structure, we have performed density functional theory (DFT) calculations, including a Hubbard U term accounting for the on-site Coulomb interaction (20). The values of U that provide a realistic description of the electronic and atomic structure of bulk YTiO3 and LaTiO3 compounds (26) were used. The atomic positions were fully relaxed, under the constraint that the in-plane lattice constant be equal to the calculated lattice constant of bulk SrTiO3. The density of electronic states, and the corresponding atomic coordinates, calculated for periodic superlattices, are shown in Fig. 4A (3.5-uc SrTiO3/1-ML LaO) and in Fig. 4B (3.5-uc SrTiO3/1-ML YO). For the LaO-based heterostructure, the Fermi energy lies in the region of nonzero density of states, consistent with the previous calculations (27, 28), whereas for the YO heterostructure the Fermi energy lies between the split-off lower Hubbard band and the higher energy density of states. This indicates that the LaO-based interface is metallic, whereas the YO-based interface is insulating, supporting our experimental observations. Our calculations predict that the ground state of the SrTiO3/LaO heterostructure is not charge-ordered, whereas the SrTiO3/YO heterostructure is unstable with respect to charge disproportionation and has a charge-ordered ground state similar to that found in (29). Octahedra rotations are clearly visible in the relaxed structures shown in Fig. 4, C and D, consistent with our synchrotron measurements.

Fig. 4

Energy-dependent density of states and structural relaxation of 3.5-uc SrTiO3/1-ML LaO (A and C) periodic superlattice and 3.5-uc SrTiO3/1-ML YO periodic superlattice (B and D) obtained from DFT calculations. Positive density of states is for spin up and negative is for spin down. The dashed line indicates the position of the Fermi level. The results indicate conducting behavior for the 3.5-uc SrTiO3/1-ML LaO periodic superlattice and insulating behavior for the 3.5-uc SrTiO3/1-ML YO periodic superlattice.

The electron donated by the RO embedded in the SrTiO3 matrix is localized to the nearby TiO2 layers. Filling of the Ti-3d band in these layers close to n = 0.5, and enhanced electron-correlation effects due to 2D confinement, will strongly influence the interfacial conductivity. It is well known that the effect of Ti-3d band filling on electronic, magnetic, and transport properties of bulk RTiO3 Mott-Hubbard insulators depends critically on the rare-earth ion (23). It appears that for the relatively weakly correlated LaO-based heterostructure, several percent of hole doping is sufficient to cause a metal-insulator transition. In contrast, for the YO-based heterostructures with larger U, lower bandwidth W, and larger strain and structural distortions, the insulating phase persists. The number of electrons transferred in each case is the same, but stronger correlation effects in the YO heterostructure seem to be responsible for the insulating behavior. Our experimental and theoretical investigations suggest that these correlations arise from an interplay of strain, spatially varying rotational distortions, and rare-earth ion effects on the band structure. Indications of electron correlations have also been recently reported in LaIO3/SrTiO3 heterostructures (30).

Strong correlations in 2DEGs at oxide interfaces have been shown to result from electronic properties of different RO inserted layers, as well as the structural and electronic modification of nearby layers. Quantitatively exploring the underlying physics of the experimental data presented here is complex and challenging, because strong correlations combined with atomic-scale structural and chemical variations severely limit the effectiveness of theoretical calculations. The details cannot be fully captured within the DFT+U calculations used here, and more advanced approaches—based on dynamical mean-field theory (31), for example—are likely necessary to capture the spatial variations. The work presented here is important in elucidating correlation effects in systems with atomic-scale perturbations (32) and external perturbation-induced changes in oxide 2DEG systems (8, 1517). The ability to design and grow heterostructures with atomic-scale variations, and the demonstrated strong dependence of correlated 2DEGs on these variations, open new directions for oxide 2DEG heterostructures.

Supporting Online Material

www.sciencemag.org/cgi/content/full/331/6019/886/DC1

Materials and Methods

Figs. S1 to S8

Table S1

References

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank D. G. Schlom and D. A. Muller for fruitful discussions. This work was supported by the National Science Foundation under grant DMR-0906443 and a David and Lucile Packard Fellowship (C.B.E.). The research at University of Nebraska was supported by the Materials Research Science and Engineering Center (NSF grant DMR-0820521), the Nanoelectronics Research Initiative of the Semiconductor Research Corporation, the National Science Foundation (grant EPS-1010674), and the Nebraska Research Initiative. Work at the University of Michigan was supported by the U.S. Department of Energy (DOE) under grant DE-FG02-07ER46416. We thank the National Center for Electron Microscopy at Lawrence Berkeley National Laboratory for their support under DOE grant DE-AC02-05CH11231 for user facilities. Work at Argonne and use of the Advanced Photon Source were supported by the DOE Office of Science, Office of Basic Energy Sciences, under contract DE-AC02-06CH11357. Work at Brookhaven National Laboratory was sponsored by DOE/BES/MSE and the Center for Functional Nanomaterials under contract DE-AC02-98CH10886. J. Karapetrova’s assistance at beamline 33-BM of the Advanced Photon Source is gratefully acknowledged.
View Abstract

Navigate This Article