High-mobility electrons in SrTiO3 heterostructures

Available online at www.sciencedirect.com

Physica E 22 (2004) 712 – 716 www.elsevier.com/locate/physe

High-mobility electrons in SrTiO3 heterostructures H.Y. Hwanga; b;∗ , A. Ohtomoc , N. Nakagawab , D.A. Mullerb , J.L. Grazulb a Department

of Advanced Materials, University of Tokyo, Kashiwa, Chiba 277-86513, Japan Laboratories, Lucent Technologies, Murray Hill, NJ 07974, USA c Institute for Materials Research, Tohoku University, Aoba, Sendai 980-8577, Japan b Bell

Abstract Recent advances in thin-4lm deposition techniques have enabled the growth of perovskite oxide heterostructures with near-atomic precision. We have been exploring the feasibility of creating two-dimensional electron gases in SrTiO3 heterostructures. Three di9erent approaches are presented here: atomic-scale delta doping, direct modulation of the oxygen stoichiometry, and mobile charge arising at a polar/non-polar heterointerface. ? 2003 Elsevier B.V. All rights reserved. PACS: 73.61.−r; 77.84.Dy; 81.15.Fg Keywords: SrTiO3 ; Electron transport

1. Introduction Perovskite oxides exhibit a broad range of physical properties—insulator, semiconductor, metal, superconductor, heavy fermion, ferromagnet, antiferromagnet, spin glass, charge/spin density wave transitions, ferroelectricity, piezoelectricity, etc. Many of these phenomena occur in materials that are lattice-matched within a few percent of one another, giving rise to the possibility of heteroepitaxial structures using perovskite oxides, accessing these multiple degrees of freedom. In this context, we have been studying SrTiO3 -based multilayer thin 4lms. SrTiO3 is a wide-band gap (Eg ∼ 3:2 eV) semiconductor that is readily doped n-type by heterovalent substitution or oxygen vacancies [1]. Despite being a fairly narrow band system (m∗ ∼ 3–5 m◦ ), the low-temperature Hall mobility of bulk-doped crystals can exceed ∗

Corresponding author. Department of Advanced Materials, University of Tokyo, Kashiwa, Chiba 277-86513, Japan. E-mail address: [email protected] (H.Y. Hwang).

10; 000 cm2 =V s, in part due to signi4cant screening of the impurity potentials by the lattice arising from a nearby ferroelectric instability [2]. As a result, SrTiO3 remains metallic for carrier densities as low as n ∼ 1017 cm−3 . In addition to being a candidate to create low-dimensional, high-mobility electron gases in oxides, there is the intriguing possibility to incorporate superconductivity at nearby carrier densities, as SrTiO3 is among the lowest density known superconductors in the range of 1019 –1021 cm−3 [3]. This aspect motivated an early exploration of 4eld e9ect devices [4], an approach that continues to be pursued [5]. After summarizing our growth techniques, here we present three di9erent experimental approaches towards forming high-mobility two-dimensional electron gases in SrTiO3 . 2. Pulsed laser deposition Pulsed laser deposition is a technique highly suited to growing multi-component oxides. A pulsed excimer

1386-9477/$ - see front matter ? 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.physe.2003.12.106

H.Y. Hwang et al. / Physica E 22 (2004) 712 – 716

713

3. Delta-doping

Fig. 1. Unit cell reLection high-energy electron di9raction (RHEED) oscillations during SrTiO3 homoepitaxial growth. One unit cell corresponds to ∼ 25 laser pulses.

laser beam (248 nm) is focused onto a solid (single crystal, polycrystalline pellet), stoichiometric target. The beam ablates an energetic plume of material, generally preserving cation stoichiometry, which is deposited on the substrate. Oxygen partial pressure is usually provided in the growth chamber. This is a broad ranging technique, addressing many di9erent materials, growth rates, etc. In the work presented here, the focus is on slow controlled deposition in the two-dimensional growth mode. Fig. 1 shows an optimized case for SrTiO3 homoepitaxy, demonstrating extended unit cell reLection high-energy electron di9raction (RHEED) oscillations. One oscillation consists of ∼ 25 laser pulses. For high-precision structures, the two-dimensional growth mode is used to directly monitor deposition on an atomic scale. Pulse-to-pulse Luctuations, long-term degradation of the transmittance of the window and the mean pulse energy, etc., dictate that the growth rate is not suMciently stable to remain at a calibrated rate.

When La3+ replaces a Sr 2+ site in a SrTiO3 host, an extra electron is introduced. In bulk, the solid solution (La; Sr)TiO3 exists all the way to LaTiO3 , corresponding to one electron per unit cell. This end member develops a Mott–Hubbard gap, resulting in an antiferromagnetic, insulating ground state [6]. Fig. 2 shows a scanning transmission electron microscopy (STEM) image of a superlattice of delta-doped structures on the atomic scale, in which complete sheets of Sr2+ in SrTiO3 have been replaced by La3+ [7]. This was grown by alternating 4ve unit cells of SrTiO3 with one unit cell of LaTiO3 . Note that to a high degree, the dopants have been con4ned to a single atomic position, with little lateral di9usion. This is despite the strong Coulomb repulsion between the charged impurities, which tends to keep them widely spaced [8]. Advantages here over delta-doping in conventional semiconductors include growth at a much lower fraction of thermodynamic temperatures and the ability to stabilize mixed valence states of Ti. The temperature-dependent resistivity and Hall mobility are given in Fig. 3, for the superlattice of Fig. 2. The as grown 4lm has extremely high conductivity and mobility, which is dominated by residual oxygen vacancies in the SrTiO3 layers (discussed in the next section). These arise during growth, because the thermodynamic conditions are compromised to attempt to stabilize both Ti4+ and Ti3+ [9]. After annealing in Lowing O2 for a few hours at 600◦ C, the residual oxygen vacancies are 4lled, and correspondingly, the resistivity increases and the mobility drops signi4cantly. The sheet density of La3+ per layer is 6:6×1014 cm−2 , and approximately 12 to 23 of the expected carriers are found to be mobile. These annealed data appear to be the intrinsic transport properties of this arti4cial superlattice. This leads to the counterintuitive conclusion that the mobility is reduced, despite the reduction in structural disorder by annealing out the oxygen vacancies. Two aspects may contribute to this e9ect: the diminished screening by reducing the carrier density, and the observation that La doped SrTiO3 has signi4cantly lower mobility than in the case of SrTiO3− for comparable densities. We believe, however, the origin of this mobility reduction has to do with the high carrier density in the vicinity of the delta-layers. Here the density peaks

714

H.Y. Hwang et al. / Physica E 22 (2004) 712 – 716

Fig. 2. Scanning transmission electron microscopy (STEM) image of a superlattice of (SrTiO3 )5 =(LaTiO3 )1 in the high angle annular dark 4eld (HAADF) imaging mode. This mode is most sensitive to atomic number, so the brightest atoms are La, the next brightest are Sr, and the Ti are weakly visible in between.

Fig. 3. The temperature-dependent resistivity (a) and Hall mobility (b) of the sample of Fig. 2, as grown and annealed at 600◦ C in Lowing O2 for a few hours.

approach those corresponding to half-4lling, and we are sampling the lower mobility of these states after annealing, after removing the parallel, higher mobility conduction channel of SrTiO3− . At low temperatures, unusual features appear in the magnetotransport properties, such as an emerging anomalous Hall e9ect. One speculative explanation is that the electron densities are suMciently high that correlation e9ects related to the Mott insulator LaTiO3 become relevant at low temperatures. The current con4guration corresponds to multiple subband occupancy, and the carrier mobility may be intrinsically limited by strong interactions. 4. Oxygen-decient thin lms A second approach to produce a controlled doping pro4le in SrTiO3 is bulk doping, either by dilute La for Sr, or Nb for Ti. In many cases for our growth conditions, the defect chemistry is non-trivial, and the free carrier density is often dominated by oxygen stoichiometry. Focusing then on SrTiO3− , the oxygen  is readily modulated by varying the growth kinetics, as controlled by the oxygen partial pressure (10−4 – 10−7 Torr) and temperature (600 –900◦ C) during growth. Fig. 4 shows STEM images of a 4lm of SrTiO ∼ 2:75 grown on SrTiO3 substrate. The interface is quite distinct between di9erent oxygen stoichiometries. There is a large kinetic barrier to oxygen di9u-

Fig. 4. STEM images of SrTiO ∼ 2:75 on SrTiO3 by HAADF imaging (a) and low angle annular dark 4eld LAADF imaging (b). LAADF is more sensitive to strain via dechanneling of the electron beam via distortions arising from vacancies.

sion, and this interface is remarkably robust despite the high growth temperatures ( ∼ 750◦ C), resulting in oxygen-doping pro4les which can be produced with nanometer abruptness. Detailed studies at high resolution reveal that the average interface width is around 0:38 nm, or one unit cell. Annealing the 4lm in 1 atm of O2 for a few hours above 400◦ C 4lls these

H.Y. Hwang et al. / Physica E 22 (2004) 712 – 716

715

Fig. 5. The temperature-dependent resistivity (a) and Hall mobility (b) of three di9erent oxygen de4cient 4lms of SrTiO3− .

vacancies formed during growth, much as in the case for the delta-doped structures described in the previous section. Each oxygen vacancy nominally “transfers” two electrons residing in the Ti d band, which are itinerant. Fine-tuning of the oxygen concentration can be used to systematically vary the carrier concentration over a wide range. Fig. 5 shows the resistivity and Hall mobility of three 100 nm thick 4lms with varying oxygen . The metal–insulator transition in 4lms is shifted to higher carrier concentrations than bulk, in part due to signi4cant surface depletion at low temperatures. This e9ect is tangible at unusually high carrier densities, enhanced considerably by the near divergence of the static dielectric constant. This depletion also provides some degree of con4nement in ultra-thin 4lms. The mobility is suMciently high that Shubnikov–de Haas magnetoresistance oscillations can be observed below a few hundred mK. Fig. 6 shows an example for a sample with a nominal room temperature carrier density of 5 × 1021 cm−3 at 100 mK, with the 4eld applied in the [0 0 1] direction. Previous studies in bulk Nb doped SrTiO3 were consistent with three ellipsoids of revolution along the 1 0 0 directions of the Brillouin zone at the X point [10]. Coincidentally, the principle feature in Fig. 6 corresponds to approximately the same Fermi-surface cross-sectional area as in Ref. [10], although the carrier density was two orders of magnitude smaller. It is diMcult to draw many conclusions from this preliminary work, because at this high carrier density, we are far from the approximation of 4lling parabolic conduction band

Fig. 6. Shubnikov–de Haas oscillations observed in SrTiO3− 4lm at 100 mK.

minima–indeed, we suspect that the nearby minima at  has been 4lled, and these various pockets have merged at these metallic densities. A complete rotation study should establish the Fermiology at these higher densities. 5. Polar/non-polar heterointerfaces The 4nal approach presented is the use of a polar/non-polar interface to present free carriers in SrTiO3 . This issue arises in many heteroepitaxial systems, and an early consideration occurred in the growth of GaAs on (0 0 1) oriented Ge [11]. Both semiconductors have the same structure and similar lattice constant, but at the interface, one is confronted with dangling bonds at the termination of the Group IV Ge layer and initiation of III–V alternations of GaAs. A similar situation can occur at the interface between LaAlO3 and SrTiO3 . LaAlO3 has a band gap in excess of 5 eV, and can be considered in the ionic extreme as alternating stacks of [La3+ O2− ]+ and [Al3+ (O2− )2 ]− in the (0 0 1) direction, as opposed to [Sr 2+ O2− ] and [Ti4+ (O2− )2 ] in SrTiO3 (Fig. 7). Key to experimentally realizing this interface is the ability to control the surface termination layer atomically [12].

716

H.Y. Hwang et al. / Physica E 22 (2004) 712 – 716

AlO2 LaO

heterointerface. This result is promising given the large electron con4ning potential ( ∼ 2 eV) provided by LaAlO3 .

AlO2 LaO TiO2 SrO TiO2 SrO Fig. 7. Schematic of the polar/non-polar interface that arises between LaAlO3 above and SrTiO3 below.

6. Summary and outlook Our current challenge is to gain suMcient control at low densities and in con4ned geometries to achieve a low subband occupation, while maintaining a high mobility. Optimistically, there appear to be no fundamental limitations, although signi4cant challenges remain. The goal is not simply to reproduce quantum Hall e9ects in an oxide host, which may have already been observed [13], but to introduce a Cooper channel at similar densities. This may provide an opportunity to examine two-dimensional superconductivity at low carrier densities. References

Fig. 8. The temperature-dependent resistivity (a) and Hall mobility (b) of the interface shown in Fig. 7.

By growing LaAlO3 on TiO2 -terminated SrTiO3 , an electron-rich interface has been obtained. A native charge assignment would give half an extra electron per two-dimensional unit cell, or 3:3 × 1014 cm−2 . Fig. 8 shows the transport properties of this structure, demonstrating a high mobility despite the unusual

[1] H.P.R. Frederikse, W.R. Thurber, W.R. Hosler, Phys. Rev. 134 (1964) 442. [2] T. Sakudo, H. Unoki, Phys. Rev. Lett. 26 (1971) 851. [3] J.F. Schooley, W.R. Hosler, M.L. Cohen, Phys. Rev. Lett. 12 (1964) 474. [4] M. Gurvitch, H.L. Stormer, R.C. Dynes, J. Graybeal, Bull. Amer. Phys. Soc. 31 (1986) 438. [5] I. Pallecchi, et al., Appl. Phys. Lett. 78 (2001) 2244. [6] Y. Tokura, et al., Phys. Rev. Lett. 70 (1993) 2126. [7] A. Ohtomo, D.A. Muller, J.L. Grazul, H.Y. Hwang, Nature 419 (2002) 378. [8] M.B. Johnson, et al., Phys. Rev. Lett. 75 (1995) 1606. [9] A. Ohtomo, D.A. Muller, J.L. Grazul, H.Y. Hwang, Appl. Phys. Lett. 80 (2002) 3922. [10] H.P.R. Frederikse, W.R. Hosler, W.R. Thurber, Phys. Rev. 158 (1967) 775. [11] G.A. Baraef, J.A. Appelbaum, D.R. Hamann, J. Vac. Sci. Technol. 14 (1977) 999. [12] M. Kawasaki, et al., Science 266 (1994) 1540. [13] M. Sasaki, et al., Solid State Commun. 109 (1999) 357.