..............................................................

Artificial charge-modulationin atomic-scale perovskitetitanate superlatticesA. Ohtomo, D. A. Muller, J. L. Grazul & H. Y. Hwang

Bell Laboratories, Lucent Technologies, Murray Hill, New Jersey 07974, USA.............................................................................................................................................................................

The nature and length scales of charge screening in complexoxides are fundamental to a wide range of systems, spanningceramic voltage-dependent resistors (varistors), oxide tunneljunctions and charge ordering in mixed-valence compounds1–6.There are wide variations in the degree of charge disproportio-nation, length scale, and orientation in the mixed-valence com-pounds: these have been the subject of intense theoretical study7–

11, but little is known about the microscopic electronic structure.Here we have fabricated an idealized structure to examine theseissues by growing atomically abrupt layers of LaTi31O3

embedded in SrTi41O3. Using an atomic-scale electron beam,we have observed the spatial distribution of the extra electron onthe titanium sites. This distribution results in metallic conduc-tivity, even though the superlattice structure is based on twoinsulators. Despite the chemical abruptness of the interfaces, wefind that a minimum thickness of five LaTiO3 layers is requiredfor the centre titanium site to recover bulk-like electronic proper-ties. This represents a framework within which the short-length-scale electronic response can be probed and incorporated in thin-film oxide heterostructures.

In perovskites, charge ordering results in modulations of theelectron density in the form of planes and slabs, whereas in lower-dimensional perovskite-derived systems, charge ordering leads tostripes, or one-dimensional charge modulations. Approximationsto the first case can be realized in thin-film superlattices in which theformal valence of the transition-metal ion is varied. Superlattices ofSrTiO3 and LaTiO3 are addressed here, where the titanium valence isvaried from 4þ to 3þ. SrTiO3 is a band insulator with an empty dband, whereas LaTiO3 has one d electron per site, and strongCoulomb repulsion results in a Mott–Hubbard insulator12. Super-lattices of these two perovskites capture many of the importantaspects of naturally occurring charge-ordered systems, namelymixed-valence configurations near half-filling. The lattice constantsare relatively well matched (for SrTiO3, a o ¼ 3.91 A; LaTiO3,pseudocubic a o ¼ 3.97 A), and the continuity of the TiO6 octa-hedral lattice across the superlattice minimizes the perturbation ofthe electronic states near the chemical potential13,14. The principalgrowth issue reduces to the control of the titanium oxidation state,which we have recently addressed for bulk-like film growth15.

We grew SrTiO3/LaTiO3 superlattice films in an ultrahigh-vacuum chamber (Pascal) by pulsed laser deposition, using asingle-crystal SrTiO3 target and a polycrystalline La2Ti2O7 target.Extreme care was taken to start with atomically flat, TiO2-terminated SrTiO3 substrates, which exhibited terraces severalhundred nanometres wide, separated by 3.91-A unit cell steps asobserved by atomic force microscopy16. A KrF excimer laser with arepetition rate of 4 Hz was used for ablation, with a laser fluence atthe target surface of ,3 J cm22. The films were grown at 750 8C withan oxygen partial pressure of 1025 torr, which represented the bestcompromise for stabilizing both valence states of titanium. Oscil-lations in the unit-cell reflection high-energy electron diffractionintensity were observed throughout the growth, and were used tocalibrate the number of layers grown. After growth, the films wereannealed in flowing oxygen at 400 8C for 2–10 hours to fill residualoxygen vacancies.

Figure 1 shows the annular dark field (ADF) image of a super-lattice sample obtained by scanning transmission electronmicroscopy (JEOL 2010F) of a 30-nm-thick cross-section along asubstrate [100] zone axis. In this imaging mode, the intensity ofscattering scales with the atomic number Z as Z 1.7, so the brightestfeatures are columns of La ions, the next brightest features arecolumns of Sr ions, and the Ti ions are weakly visible in between17–19.The quality of the interfaces does not degrade with continueddeposition, and the atomic step and terrace structure of the growingsurface is maintained for hundreds of nanometres. The magnifiedview at the top of Fig. 1 shows a higher-resolution image, whichvisibly demonstrates the ability to grow a single layer of La ions.Because the layer is viewed in projection, roughness along thebeam—particularly on length scales thinner than the sample—leads to apparent broadening. Thus these results represent anupper limit to the actual width of the layers.

With the same imaging conditions used to obtain Fig. 1, weanalysed the energy of the transmitted electron beam and per-formed core level spectroscopy, atom column by atom column20–22.This approach is able to probe internal structures directly, unlikesurface-sensitive methods. Specifically, the titanium L2,3, oxygen K,and lanthanum M4,5 edges can be simultaneously recorded, with anenergy resolution of ,0.9 eV and a spatial resolution slightly worsethan the ADF resolution of ,1.9 A, primarily owing to drift duringthe slower acquisition of the spectra. We obtained a scan throughthe Ti sites crossing a 2-unit-cell layer of LaTiO3 (top centre panel ofFig. 2). By substituting La for Sr, there is locally an extra electronthat resides mainly on the Ti d orbitals23. To visualize this effect, theTi L2,3 near-edge structure can be decomposed into a linearcombination of Ti3þ and Ti4þ, with no residual detectable abovethe experimental noise level (bottom panel of Fig. 2).

This decomposition, which would fail both conceptually andexperimentally for more covalent materials, allows a particularly

Figure 1 Annular dark field (ADF) image of LaTiO3 layers (bright) of varying thickness

spaced by SrTiO3 layers. The view is down the [100] zone axis of the SrTiO3 substrate,

which is on the right. After depositing initial calibration layers, the growth sequence is

5 £ n (that is, 5 layers of SrTiO3 and n layers of LaTiO3), 20 £ n, n £ n, and finally a

LaTiO3 capping layer. The numbers in the image indicate the number of LaTiO3 unit cells

in each layer. Field of view, 400 nm. Top, a magnified view of the 5 £ 1 series. The raw

images have been convolved with a 0.05-nm-wide gaussian to reduce noise.

letters to nature

NATURE | VOL 419 | 26 SEPTEMBER 2002 | www.nature.com/nature378 © 2002 Nature Publishing Group

simple extraction of the distribution of the extra electron. The sametrend has been observed qualitatively using Ti 2p X-ray absorptionspectroscopy for bulk La12xSrxTiO3, to which this technique isformally equivalent24. This unusually ionic decomposition is prob-ably the consequence of three effects: (1) the octahedral symmetryof the Ti sites is preserved; (2) the extra electron resides on the Tisite; and (3) the Ti 2p–3d–3d couplings in the presence of the corehole are large compared with the dispersion widths of the 3dbands24. The onset of Ti3þ admixture, visible in the raw spectra,occurs in the vicinity of the double layer of LaTiO3, which issignalled by the La M5 edge. The oxygen concentration across thelayer remains unchanged within experimental error.

We also show the result of a similar scan across a single layer ofLaTiO3 (Fig. 3a). The spread of the extra electron can be seen in thefractional Ti3þ signal, which has a full-width at half-maximum of19 ^ 2 A. Because this is significantly wider than the integrated LaM4,5 edge, the extended distribution of the extra electron can beconsidered intrinsic to the ideal structure, not limited by interfaceroughness or disorder. Figure 3b shows that the Ti3þ signal for boththe single- and double-layer LaTiO3 structures can be fitted with anexponential ‘tail’, giving a decay length of l ¼ 1.0 ^ 0.2 nm. As an

order-of-magnitude estimate of the expected screening length, weconsider the Thomas–Fermi approximation for a degenerate semi-conductor, assuming unity dopant activation. For the range ofreported values for bulk and thin film SrTiO3 (10 /m* varying from10 to 100, where 10 is the static dielectric constant and m* is theeffective mass), a screening length in the range 0.23–0.72 nm isdeduced. This approximate correspondence with the experimentalresult is remarkable, given the borderline applicability of theeffective mass approximation, the atomically abrupt potentialvariation, and the lack of consideration of the phonon structureof 1(q), the momentum-dependent dielectric response.

In Fig. 4a we compare the number of LaTiO3 unit cells in a layerwith the integrated Ti3þ area across the layer—the expected linearrelationship may be seen. Note that the measured area is calibratedin an absolute sense, with the Ti3þ area observed per LaTiO3 layeralmost completely (95 ^ 3%) accounting for the extra electron.More importantly, no evidence for compensating cation defects wasobserved. Figure 4b displays the peak fraction of Ti3þ as a functionof the number of LaTiO3 unit cells. An extrapolation of the existingdata shows that 5 unit cells are needed for the central Ti site toexhibit bulk-like spectroscopic features.

In much of the region associated with the LaTiO3 layers, the Tisites exhibit mixed valence between 3þ and 4þ. In bulk solidsolutions of La12xSrxTiO3, this configuration of the Ti valenceresults in metallic behaviour, and correspondingly, superlattices ofLaTiO3/SrTiO3 are metallic. The overall conductivity is a functionof the spacing between, and the thickness of, the LaTiO3 interlayers.

Figure 2 Electron energy-loss spectra (EELS) for La and Ti simultaneously recorded

across a 2-unit-cell LaTiO3 layer in SrTiO3. Left, the ADF image for the layer (layer ‘2’ in

Fig. 1); centre, the experimental Ti L2,3 edge (grey lines), with the left side of the spectra

aligned to the ADF image. The electron beam position for these data is denoted by the

dashed line in the ADF image. Reference spectra for Ti4þ (red) and Ti3þ (blue) are shown

at the bottom, taken from thick sections of SrTiO3 and LaTiO3. Coloured lines show fits to

the position-dependent spectra colour-coded by the fractional contribution of Ti4þ and

Ti3þ. Right, the La M5 edge, aligned to the ADF image. Note that scanning along the Ti

sites broadens the La signal beyond the intrinsic resolution function. Bottom panel,

detailed view of the decomposition of the Ti L2,3 edge for the Ti site in the middle of the

layer. Experimental data are shown as black dots, fitted (violet) by the addition of the

reference spectra shown in red (Ti4þ) and blue (Ti3þ). The residual to the fit is given by the

black line at the bottom.

Figure 3 Spatial distribution of the Ti3þ signal in the vicinity of the LaTiO3 layer and

bilayer. a, EELS profiles for La and Ti recorded across a LaTiO3 monolayer. Inset, the ADF

image for the monolayer (layer ‘1’ in Fig. 1). The La M edge is recorded simultaneously

with the Ti L edge, yet the Ti3þ signal is considerably wider than that of the La. The

absolute fractions of La and Ti3þ were calibrated from bulk LaTiO3 and SrTiO3. b, The

decay of the Ti3þ signal away from the LaTiO3 monolayer of a as well as the bilayer of

Fig. 2. The tails of the Ti3þ signal for both structures fit an exponential decay with a decay

length of l ¼ 1.0 ^ 0.2 nm.

letters to nature

NATURE | VOL 419 | 26 SEPTEMBER 2002 | www.nature.com/nature 379© 2002 Nature Publishing Group

By examining the Hall effect in a number of superlattice compo-sitions, we can measure the effective free carrier density n, which isnearly temperature independent ðn¼21=RHe; where RH is theHall coefficient and e is the electron charge). For a wide range ofcompositions, the measured carrier density corresponds to roughly2/3 of the carriers expected from the La fraction (Fig. 4c).

In the artificial charge-modulated structures described here, themeasured electronic structure reflects the static equilibrium distri-bution, including screening effects from the lattice. Although thecharge modulation that we observe is driven by the presence ofcharged donor LaO layers, and not by spontaneous valence frac-tionation driven by strong correlation effects, the electronicresponse probed by this perturbation should be relevant for staticcharge ordering observed in bulk perovskites. Theoreticalapproaches addressing charge ordering range from those focusingon nearest-neighbour interactions in Hubbard and t–J modelsneglecting long-range Coulomb interactions, to those which requirelong-range interactions to stabilize inhomogeneous charge distri-butions7,8,10. As a function of the relative strength of the localinteractions, solutions vary from abrupt solitonic domain lines, tosmoothly varying domain walls approaching conventional densitywave structures9. The abrupt, short-length-scale features deducedfrom naturally occurring charge-ordered systems6 were notobserved in our measurements. Indeed, in superlattice structurescloser to half-filling, such as occasional SrTiO3 layers withinLaTiO3, we see no short-length-scale structure, but rather a smoothaverage electron distribution (data not shown). Whether this is aconsequence of the specific electronic parameters in the titanates, ora result of constraints arising from our artificial structure, remainsto be understood. A

Received 9 April; accepted 12 July 2002; doi:10.1038/nature00977.

1. Greuter, F. & Blatter, G. Electrical properties of grain boundaries in polycrystalline compound

semiconductors. Semicond. Sci. Technol. 5, 111–137 (1990).

2. Chaudhari, P. et al. Direct measurement of the superconducting properties of single grain boundaries

in Y1Ba2Cu3O7-d. Phys. Rev. Lett. 60, 1653–1656 (1988).

3. Sun, J. Z. et al. Observation of large low-field magnetoresistance in trilayer perpendicular transport

devices made using doped manganate perovskites. Appl. Phys. Lett. 69, 3266–3268 (1996).

4. Chen, C. H., Cheong, S-W. & Cooper, A. S. Charge modulations in La2-xSrxNiO4þy: ordering of

polarons. Phys. Rev. Lett. 71, 2461–2464 (1993).

5. Tranquada, J. M., Sternlieb, B. J., Axe, J. D., Nakamura, Y. & Uchida, S. Evidence for stripe correlations

of spins and holes in copper oxide superconductors. Nature 375, 561–563 (1995).

6. Mori, S., Chen, C. H. & Cheong, S-W. Pairing of charge-ordered stripes in (La,Ca)MnO3. Nature 392,

473–476 (1998).

7. Zaanen, J. & Gunnarsson, O. Charged magnetic domain lines and the magnetism of high-T c oxides.

Phys. Rev. B 40, 7391–7394 (1989).

8. Poilblanc, D. & Rice, T. M. Charged solitons in the Hartree-Fock approximation to the large-U

Hubbard model. Phys. Rev. B 39, 9749–9752 (1989).

9. Inui, M. & Littlewood, P. B. Hartree-Fock study of the magnetism in the single-band Hubbard model.

Phys. Rev. B 44, 4415–4422 (1991).

10. Loew, U., Emery, V. J., Fabricius, K. & Kivelson, S. A. Study of an Ising model with competing long-

and short-range interactions. Phys. Rev. Lett. 72, 1918–1921 (1994).

11. Kivelson, S. A., Fradkin, E. & Emery, V. J. Electronic liquid-crystal phases of a doped Mott insulator.

Nature 393, 550–553 (1998).

12. Tokura, Y. et al. Filling dependence of electronic properties on the verge of metal-Mott-insulator

transitions in Sr1-xLaxTiO3. Phys. Rev. Lett. 70, 2126–2129 (1993).

13. Sunstrom, J. E. IV, Kauzlarich, S. M. & Klavins, P. Synthesis, structure, and properties of La1-xSrxTiO3

(0 # x # 1). Chem. Mater. 4, 346–353 (1992).

14. Kahn, A. H. & Leyendecker, A. J. Electronic energy bands in strontium titanate. Phys. Rev. 135,

A1321–A1325 (1964).

15. Ohtomo, A., Muller, D. A., Grazul, J. L. & Hwang, H. Y. Epitaxial growth and electronic structure of

LaTiOx films. Appl. Phys. Lett. 80, 3922–3924 (2002).

16. Kawasaki, M. et al. Atomic control of the SrTiO3 crystal surface. Science 266, 1540–1542 (1994).

17. Howie, A. Image contrast and localized signal selection techniques. J. Microsc. 17, 11–23 (1979).

18. Pennycook, S. J. Z contrast STEM for materials science. Ultramicroscopy 30, 58–69 (1989).

19. Kirkland, E. J., Loane, R. F. & Silcox, J. Simulation of annular dark field STEM images using a modified

multislice method. Ultramicroscopy 23, 77–96 (1987).

20. Batson, P. E. Simultaneous STEM imaging and electron energy-loss spectroscopy with atomic column

sensitivity. Nature 366, 727–728 (1993).

21. Muller, D. A., Tzou, Y., Raj, R. & Silcox, J. Mapping sp2 and sp3 states of carbon at sub-nanometre

spatial resolution. Nature 366, 725–727 (1993).

22. Browning, N. D., Chisholm, M. M. & Pennycook, S. J. Atomic-resolution chemical analysis using a

scanning transmission electron microscope. Nature 366, 143–146 (1993).

23. Zaanen, J., Sawatzky, G. A. & Allen, J. W. Band gaps and electronic structure of transition-metal

compounds. Phys. Rev. Lett. 55, 418–421 (1985).

24. Abbate, M. et al. Soft-x-ray-absorption studies of the location of extra charges induced by substitution

in controlled-valence materials. Phys. Rev. B 44, 5419–5422 (1991).

25. Taguchi, Y. et al. Critical behaviour in LaTiO3þd/2 in the vicinity of antiferromagnetic instability. Phys.

Rev. B 59, 7917–7924 (1999).

AcknowledgementsWe thank G. E. Blumberg, R. de Picciotto, B. I. Halperin, D. R. Hamann, S. H. Simon,C. M. Varma and N. Zhitenev for discussions. A.O. acknowledges partial support by theNishina Memorial Foundation, Japan.

Competing interests statementThe authors declare that they have no competing financial interests.

Correspondence and requests for materials should be addressed to H.Y.H.

(e-mail: hyhwang@lucent.com).

Figure 4 Summary of the electronic properties in SrTiO3/LaTiO3 superlattices as a

function of the number of LaTiO3 unit cells. a, The Ti3þ integrated area across LaTiO3

layers 1–4 unit cells thick, as calibrated from bulk LaTiO3. b, The Ti3þ peak fraction

across these layers, extrapolating to 100% for 5-unit-cell-thick LaTiO3. c, The carrier

density from Hall effect measurements for various superlattices of [(SrTiO3)m /

(LaTiO3)n ]10, compared with bulk LaxSr12xTiO3 from refs 12 and 25. The dashed line

represents 2/3 of the La contributing free carriers.

letters to nature

NATURE | VOL 419 | 26 SEPTEMBER 2002 | www.nature.com/nature380 © 2002 Nature Publishing Group