LETTERSPUBLISHED ONLINE: 21 JUNE 2009 | DOI: 10.1038/NMAT2486

Atomic-scale imaging of individual dopant atomsin a buried interfaceN. Shibata1,2*, S. D. Findlay1, S. Azuma1, T. Mizoguchi1, T. Yamamoto1,3 and Y. Ikuhara1,3,4

Determining the atomic structure of internal interfaces inmaterials and devices is critical to understanding theirfunctional properties. Interfacial doping is one promisingtechnique for controlling interfacial properties at the atomicscale1–5, but it is still a major challenge to directly characterizeindividual dopant atoms within buried crystalline interfaces.Here, we demonstrate atomic-scale plan-view observation ofa buried crystalline interface (an yttrium-doped alumina high-angle grain boundary) using aberration-corrected Z-contrastscanning transmission electron microscopy. The focusedelectron beam transmitted through the off-axis crystalsclearly highlights the individual yttrium atoms located on themonoatomic layer interface plane. Not only is their uniquetwo-dimensional ordered positioning directly revealed withatomic precision, but local disordering at the single-atom level,which has never been detected by the conventional approaches,is also uncovered. The ability to directly probe individual atomswithin buried interface structures adds new dimensions to theatomic-scale characterization of internal interfaces and otherdefect structures inmany advancedmaterials and devices.

Controlling crystal interfaces at the atomic scale is now activelybeing sought in the broader range of materials science and deviceengineering fields. One underlying motivation is that interfaces,owing to the abrupt structural and chemical inhomogeneity inthe localized volume, may exhibit novel functional properties thatcannot be realized in bulk crystals. In recent years, changing thelocal chemistry of an interface by doping foreign atoms has beenshown to markedly modify the atomic and electronic structure ofthe interface and hence its properties1–7. However, understandingthe fundamental role of the interfacial dopants is still a non-trivialtask because it is extremely difficult to characterize individualdopant atomswithin the two-dimensional defects insidematerials.

A popular method for characterizing internal interfaces isatomic-resolution transmission electron microscopy8 (TEM). Inthis method, interfaces are generally observed from a low-indexcrystallographic direction parallel to the interface (so-called cross-sectional imaging), with the interface atoms imaged in projectionalong the viewing direction. In recent years, Z -contrast scanningTEM (STEM; ref. 9) has become capable of directly imaging dopantatoms within the bulk10 and interfaces3–5,11,12 (the Z in Z -contrastdenotes the atomic number). One may extract some informationon the three-dimensional dopant atom positioning by combiningSTEM observations from two orthogonal directions parallel to theinterface plane13, but rigorous determination of individual dopantatom positions by the projected images is in principle extremelydifficult. In crystalline materials, strong electron channelling alongatomic columns enhances imaging of the crystal structure, but may

1Institute of Engineering Innovation, School of Engineering, The University of Tokyo, Bunkyo, Tokyo 113-8656, Japan, 2PRESTO, Japan Science andTechnology Agency, 4-1-8 Honcho Kawaguchi, Saitama 332-0012, Japan, 3Nanostructures Research Laboratory, Japan Fine Ceramics Center, 2-4-1Mutsuno, Atsuta, Nagoya 456-8587, Japan, 4WPI Advanced Institute for Materials Research, Tohoku University, 2-1-1, Katahira, Aoba, Sendai 980-8577,Japan. *e-mail: shibata@sigma.t.u-tokyo.ac.jp.

make determining the dopant atom positions along the atomiccolumns impossible14. On the other hand, the three-dimensionalpositioning of individual dopant atoms within an amorphousmatrix can be directly probed by STEM depth sectioning15, becausewithout strong electron channelling, one may more reliably focusthe electron beam on single dopant atoms inside the matrix,maximizing their contribution while minimizing the contributionfrom the surroundings. Even in crystalline materials, it has beenpredicted that dopant atom positioning along the depth directioncan be determined if the imaging is carried out from off-axis orweak channelling crystallographic orientations14. Thus, it should bepossible to selectively highlight individual dopant atoms within aburied crystalline interface if the electron beam probes the interfaceplane under very weak electron channelling conditions.

Here, we demonstrate Z -contrast STEM observation of anyttrium (Y)-doped grain boundary in alumina (α-Al2O3) from thedirection perpendicular to the interface plane. In these observa-tions, we focused a very fine electron beam onto the interface planethrough the off-axis α-Al2O3 crystals, to probe and detect the indi-vidual Y atoms within the interface. Y atoms are known tomarkedlystrengthen theα-Al2O3 interface againstmechanical deformation16,and their detailed atomic-scale positioning is thought to be ofprimal importance to understand the strengthening mechanism4.We first fabricated a model Y-doped α-Al2O3 grain boundary bydiffusion bonding of two single crystals in the 6 13 orientationrelationship17 (the 6 value represents the degree of geometricalcoincidence at the grain boundary). Y was added to the interfacebefore diffusion bonding, as described elsewhere16. Figure 1a showsschematically the bicrystal fabricated in this study. The bicrystallog-raphy of the interface between the top and bottom crystals is sum-marized as follows: (101̄4)top ‖ (101̄4)bottom, [12̄10]top ‖ [1̄21̄0]bottomand [2̄021]top ‖ [2̄021]bottom. The interface normal is parallel tothe high-index 〈505̄4〉 directions in both the top and bottomcrystals. Figure 1b shows typical atomic-resolution Z -contrastSTEM images of the interface projected along the 〈12̄10〉 and 〈2̄021〉directions. The doped Y atomic columns are clearly imaged withvery strong contrast along the boundary, and form a monoatomiclayer structure in the core of the boundary. Slight lattice distortioncan be seen in the 〈12̄10〉 projected image, suggesting the presenceof local strain in the vicinity of the boundary core region.

Figure 2a shows a low-magnification plan-view bright-fieldTEM image of the Y-doped boundary observed from the 〈505̄4〉direction. The inset shows the corresponding electron diffractionpattern obtained from the plan-view interface area. Diffractionanalysis confirmed that the 〈505̄4〉 direction is about 6◦ offfrom the nearest low-index 〈101̄1〉 directions in both the topand bottom crystals. The bright-field image shows a periodic

654 NATUREMATERIALS | VOL 8 | AUGUST 2009 | www.nature.com/naturematerials

© 2009 Macmillan Publishers Limited. All rights reserved.

NATUREMATERIALS DOI: 10.1038/NMAT2486 LETTERS

[1210]

[0001]

[0001]

[2021]

[5054]

[5054]

[2021]

[1210]

¬

¬

¬

¬

¬ ¬

¬

Y 103.6°Grain boundary

-Al2O3

1210¬

2021¬

1 nm

1 nm

a

b

α

-Al2O3α

Figure 1 | Schematic illustration and two cross-sectional Z-contrastSTEM images of the Y-doped6 13 grain boundary of α-Al2O3.a, Schematic illustration of the α-Al2O3 bicrystal fabricated in this study,with the orientation relationship between the top and bottom crystalsindicated. Y atoms are artificially doped in the boundary plane.b, Z-contrast STEM images of the Y-doped grain boundary observed fromtwo orthogonal directions parallel to the interface plane. Bright spots in thegrain interiors directly correspond to the position of the Al atomic columns.The doped Y atomic columns are clearly imaged with very strong contrastand form a monoatomic layer structure in the core of the boundary.

network of dislocation contrast throughout the plan-view interface.These network dislocations are secondary interfacial dislocations18introduced to compensate for a slightmisorientation from the exact6 13 orientation. This is a clear indication that the interface plane

¬a

b

⟨2021⟩

¬⟨2021⟩

¬⟨1210⟩

¬⟨1210⟩

¬⟨5054⟩

¬⟨5054⟩1 μm

1 nm

Figure 2 | Plan-view images of the Y-doped6 13 grain boundary. a, Abright-field TEM image of the Y-doped6 13grain boundary observed fromthe 〈505̄4〉 direction perpendicular to the interface plane. The inset showsthe electron diffraction pattern obtained from the observed area. Periodicnetwork dislocation contrast is clearly seen in the image, indicating thepresence of the interface within the observed volume. b, A Z-contrastSTEM image of the Y-doped6 13 grain boundary observed from the〈505̄4〉 direction. The image was obtained from the region in between thenetwork dislocations. The interface Y atoms are visible as strong imageintensity spots on weak background contrast elongated along the 〈2̄021〉direction. The STEM image was smoothed by a smoothing filterincorporated in DigitalMicrograph, Gatan.

is indeed preserved within the observed volume of the thin TEMspecimen. In the following atomic-scale observations, we focus onthe area in between these network dislocations, which contains themonoatomic layer grain-boundary structure.

Figure 2b shows a high-resolution Z -contrast STEM imageof the Y-doped 6 13grain boundary observed from the 〈505̄4〉direction perpendicular to the interface plane. In this plan-viewimage, the interface Y atoms are visible as strong image intensityspots owing to their much larger atomic number (Z ) comparedwith Al and O atoms. These spots are periodically arrayed alongthe 〈2̄021〉 direction, and accompany a weak background contrastelongated along the same direction. This striped backgroundcorresponds to the projected image of Al atomic planes of thematrix α-Al2O3, as will be shown later. It is found that the Yatom occupancy is not uniform on these stripes; Y atoms are

NATUREMATERIALS | VOL 8 | AUGUST 2009 | www.nature.com/naturematerials 655© 2009 Macmillan Publishers Limited. All rights reserved.

LETTERS NATUREMATERIALS DOI: 10.1038/NMAT2486

Grain boundary

5054¬

5054¬

1210¬

1210¬

2021¬

2021¬Al O

50

¬1 0x position (Å) x position (Å)

Pristine Y-doped

1 ¬1 0 1

100

150

200

250

300

350

Dep

th (

Å)

a b

c

0.020.040.060.080.100.120.14

Vacuum Interface

Figure 3 | Z-contrast STEM image simulation of the plan-view Y-doped6 13 grain boundary. a, Atomic structure model of the pristine6 13 grainboundary used for the STEM image simulations. To model the Y-doped interface, the Al atom indicated by the green circle in the boundary core is replacedwith Y. b, The simulated intensity profiles of the electron beam, projected along the 〈12̄10〉 direction, in complete vacuum and focused on the interfaceplane between the crystals. The position of the interface plane is shown by the dashed line. The thickness of the sample is assumed to be 355 Å,approximately that estimated by STEM electron energy-loss spectroscopy. c, The simulated Z-contrast STEM images of the pristine and Y-doped6 13grainboundary observed from the 〈505̄4〉 direction. Corresponding structure models are shown above the images. In the pristine case, striped contrast along the〈2̄021〉 direction due to the periodic Al-containing atomic planes is seen. In the Y-doped case, in addition to the stripes we further see the bright contrastspot corresponding to the interface Y atom.

much more densely situated on every second stripe. These resultsindicate that Y atoms preferentially occupy specific atomic siteson the interface plane. The plan-view image thus directly revealsthe two-dimensional array of Y atoms on the buried interfaceplane. To theoretically support the visibility of the Y atoms inthe 6 13 α-Al2O3 grain boundary, we carried out multisliceSTEM image simulations19. Figure 3a shows a schematic diagramof the model pristine 6 13 α-Al2O3 grain boundary used in thesimulations. This structure is the stable grain-boundary structurepredicted by first-principles calculations, and is consistent withprevious experiments17,20,21. Figure 3b shows simulated intensityprofiles, projected along the 〈12̄10〉 direction, of the electronbeam in complete vacuum and focused on the interface planebetween the crystals. The intensity profiles are almost identical,suggesting the atomic-resolution capability of the probe in vacuumis preserved inside the α-Al2O3 crystal. Figure 3c shows simulatedplan-view images of pristine and Y-doped 6 13 grain boundaries.Corresponding structure models viewed from the 〈505̄4〉 directionare also shown, and the Y position is indicated. In the Y-doped case,we have simply substituted a single Y atom into the grain-boundaryAl site as indicated in Fig. 3a. Preliminary density functionaltheory calculations support such substitution as being energeticallyfavourable, as seen previously in a similar system4. Structuralrelaxation due to the Y substitution was not included in the presentsimulation, but the small displacements involved would not change

our conclusions about the visibility of the Y atoms. In the pristinecase, the periodic striped contrast parallel to the 〈2̄021〉 direction isdue to the periodic Al-containing atomic planes. In the Y-dopedcase, in addition to the stripes due to the Al-containing atomicplanes, we further see the bright contrast spot corresponding tothe interface Y atom. This result is in good agreement with ourexperimental images. The thickness dependence of the Y contrastis not as large as might be expected in lower-order zone axis cases,as shown in the Supplementary Information, and the Y atom isclearly visible even for thismoderately thick specimen.We concludethat atomic-scale observation on the buried crystalline interfaceis possible under the present observing conditions, and that theexperimentally observed strong spots on the striped backgroundcontrast are indeed single Y atoms on the interface plane.

As the striped background originates from the Al atomic planesin the matrix α-Al2O3 crystals, we carried out image filtering toremove it and so improve the visibility of the Y atom arrangement.We used a simple fast Fourier transformmaskingmethod to removethe background stripes from the original image. The detailedprocedure and a comparison with another filtering technique aregiven in the Supplementary Information. Figure 4 shows the filteredversion of the image in Fig. 2b. The two-dimensional array ofindividual Y atoms is nowmore clearly evident. It is obvious that theY atoms are well ordered in two dimensions, selectively occupyingspecific atomic sites on the interface plane. We have checked that

656 NATUREMATERIALS | VOL 8 | AUGUST 2009 | www.nature.com/naturematerials

© 2009 Macmillan Publishers Limited. All rights reserved.

NATUREMATERIALS DOI: 10.1038/NMAT2486 LETTERS

1 nm

5054¬

1210¬

2021¬

Figure 4 | Filtered plan-view Z-contrast STEM image highlighting thetwo-dimensional positioning of the interface Y atoms. The filtered imagewas obtained by a simple fast Fourier transform masking method to removethe background stripes from the original image shown in Fig. 2b. The imageis shown with a nonlinear intensity scale to highlight bright features. Thetwo-dimensional ordered array of individual Y atoms is now more clearlyevident. The comparison of Y atom positioning between the filtered imageand the two cross-sectional images, obtained from two separate TEMsamples prepared from the same initial bicrystal, shows that the Y atompositions are in excellent agreement when observed from these differentimaging directions. However, we sometimes see Y atoms in between theordered array of Y atoms in the plan-view image, as indicated by the redarrows. These atoms are not generally seen in cross-sectional imagesowing to their very low atomic density along the viewing directions.

the subtraction of the stripes does not shift the apparent peaklocations, although without the stripes as a guide to the eye thesmall displacements from the ideally ordered array seem morepronounced in Fig. 4. The comparison of the Y atom positioningbetween the filtered plan-view image and the two cross-sectionalimages, obtained from two separate TEM samples prepared fromthe same initial bicrystal, is also shown in the figure. The Y atomarray along the 〈12̄10〉 and 〈2̄021〉 directions corresponds well tothe Y atomic column positions observed in the two cross-sectionalimages. In addition, there are some clear deviations of the Y atompositioning from the ordered positions. We sometimes see Y atomsin between the ordered array sites, as indicated by the arrows.These atoms are not common features, but were routinely found inextensive observations from different areas. The present results thusdemonstrate that Y positioning at the grain boundary is basicallyordered in two dimensions based on specific stable atom positions

on the interface, but that some disordering exists, perhaps owingto trapping and/or overflowing onto the metastable sites. The twocross-sectional images shown in Fig. 1b could not detect these strayY atoms because of their very low density along the projecteddirections. Thus, the presence of locally disordered interface atomshas been overlooked for years by conventional cross-sectional TEMobservations. These disordered interface atoms correspond to theatomic-scale quality/roughness of interfaces, which may have astrong influence on their functional properties.

Our study clearly demonstrates that Z -contrast STEM withoff-axis illumination can be a very powerful method for directlyimaging individual atoms within buried crystalline interfaces,bringing us a crucial step towards the full three-dimensionalcharacterization of interface atomic structures inside materials.It is anticipated that the next generation of STEM aberrationcorrectors will further increase the visibility of the interface atoms,improving the lateral and depth resolution and the sensitivity22–24.The possibility of directly analysing buried atomic structures willsubstantially assist the development of advanced materials anddevices engineered at the atomic scale.

MethodsAn Y-doped6 13 α-Al2O3 bicrystal was fabricated by diffusion bonding at 1,500 ◦Cfor 10 h in air. The detailed procedure, including the Y addition, has been reportedelsewhere16. The specimens for STEM observations were prepared as follows. Thebicrystal was cut and mechanically polished with a diamond suspension to havea total thickness of about 10–50 µm. In the Ar ion-beam thinning processes, weused an initial accelerating voltage of 4.0 kV and an incident beam angle of around7◦–10◦. The accelerating voltage was gradually decreased to 1.0 kV as thinningprogressed. Special care was taken when fabricating the plan-view specimen toretain the interface within the thinned volume. Bright-field TEM images weretaken with a 200 kV JEM-2010HC electron microscope (JEOL). Atomic-resolutionZ -contrast STEM images were taken with a 200 kV JEM-2100F TEM/STEMelectron microscope (JEOL) equipped with an aberration corrector (CEOSGmbH), providing a minimum probe diameter of about 1 Å. In the Z -contrastimaging, a probe convergence angle of 27.4mrad and an annular dark-fielddetector with an inner angle greater than 80mrad were used. Z -contrast STEMsimulations were carried out in a multislice model based on an effective inelasticscattering potential19. The 200 keV probe was assumed to be aberration-free, with aprobe convergence angle of 27.4mrad. The annular detector was assumed to spanthe range 81–228mrad. The simulations assumed the grain boundary to be in thespecimen mid-plane and a defocus value that placed the beam waist at the grainboundary. Finite effective source size, characterized by a Gaussian distributionof half-width at half-maximum 0.4 Å, was included in the plan-view simulations.Debye–Waller factors were taken from the literature25. The Debye–Waller factorfor the Y atom was taken to be identical to that of the Al atoms. Simulations showedlittle variation in Y-dopant visibility with reasonable variations of this value.

Received 13 February 2009; accepted 19 May 2009;published online 21 June 2009

References1. Hammerl, G. et al. Enhanced supercurrent density in polycrystalline

YBa2Cu3O7−δ at 77 K from calcium doping of grain boundaries. Nature 407,162–164 (2000).

2. Sorba, L. et al. Tuning AlAs-GaAs band discontinuities and the role ofSi-induced local interface dipoles. Phys. Rev. B. 43, 2450–2453 (1991).

3. Shibata, N. et al. Observation of rare-earth segregation in silicon nitrideceramics at subnanometre dimensions. Nature 428, 730–733 (2004).

4. Buban, J. P. et al. Grain boundary strengthening in alumina by rare earthimpurities. Science 311, 212–215 (2006).

5. Sato, Y. et al. Role of Pr segregation in acceptor-state formation at ZnO grainboundaries. Phys. Rev. Lett. 97, 106802 (2006).

6. Luzzi, D. E., Yan, M., Šob, M. & Vitek, V. Atomic structure of a grain boundaryin a metallic alloy: Combined electron microscope and theoretical study.Phys. Rev. Lett. 67, 1894–1897 (1991).

7. Muller, D. A. & Mills, M. J. Electron microscopy: Probing the atomic structureand chemistry of grain boundaries, interfaces and defects. Mater. Sci. Eng. A260, 12–28 (1999).

8. Williams, D. B. & Carter, C. B. Transmission Electron Microscopy: A Text Bookfor Materials Science. Imaging III (Plenum, 1996).

9. Pennycook, S. J. & Jesson, D. E. High-resolution Z -contrast imaging of crystals.Ultramicroscopy 37, 14–38 (1991).

10. Voyles, P. M. et al. Atomic-scale imaging of individual dopant atoms andclusters in highly n-type bulk Si. Nature 416, 826–829 (2002).

NATUREMATERIALS | VOL 8 | AUGUST 2009 | www.nature.com/naturematerials 657© 2009 Macmillan Publishers Limited. All rights reserved.

LETTERS NATUREMATERIALS DOI: 10.1038/NMAT2486

11. Chisholm, M. F., Maiti, A., Pennycook, S. J. & Pantelides, S. T. Atomicconfigurations and energetics of arsenic impurities in a silicon grain boundary.Phys. Rev. Lett. 81, 132–135 (1998).

12. Yan, Y. et al. Impurity-induced structural transformation of a MgO grainboundary. Phys. Rev. Lett. 81, 3675–3678 (1998).

13. Winkelman, G. B. et al. Three-dimensional organization of rare-earth atoms atgrain boundaries in silicon nitride. Appl. Phys. Lett. 87, 061911 (2005).

14. Xin, H. L., Intaraprasonk, V. & Muller, D. A. Depth sectioning of individualdopant atoms with aberration-corrected scanning transmission electronmicroscopy. Appl. Phys. Lett. 92, 013125 (2008).

15. van Benthem, K. et al. Three-dimensional imaging of individual hafniumatoms inside a semiconductor device. Appl. Phys. Lett. 87, 034104 (2005).

16. Matsunaga, K. et al. Direct measurements of grain boundary sliding inyttrium-doped alumina bicrystals. Appl. Phys. Lett. 82, 1179–1181 (2003).

17. Fabris, S. & Elsässer, C. 6 13 (101̄4) twin in α-Al2O3: A model for a generalgrain boundary. Phys. Rev. B 64, 245117 (2001).

18. Sutton, A. P. & Balluffi, R. W. Interfaces in Crystalline Materials (Oxford Univ.Press, 1995).

19. Allen, L. J., Findlay, S. D., Oxley, M. P. & Rossouw, C. J. Ultramicroscopy 96,47–63 (2003).

20. Nakamura, K. et al. First-principles study of grain boundary sliding in α-Al2O3.Phys. Rev. B 75, 184109 (2007).

21. Höche, T. et al. The structure of special grain boundaries in α-Al2O3.J. Phys. Chem. Solids 55, 1067–1082 (1994).

22. Borisevich, A. Y., Lupini, A. R. & Pennycook, S. J. Depth sectioning with theaberration-corrected scanning transmission electron microscope. Proc. NatlAcad. Sci. USA 103, 3044–3048 (2006).

23. Haider, M. et al. Prerequisites for a Cc/Cs-corrected ultrahigh-resolution TEM.Ultramicroscopy 108, 167–178 (2008).

24. Krivanek, O. L. et al. An electron microscope for the aberration-corrected era.Ultramicroscopy 108, 179–195 (2008).

25. Kirfel, A. & Eichhorn, K. Accurate structure analysis with synchrotronradiation. The electron density in Al2O3 and Cu2O. Acta Crystallogr. A 46,271–284 (1990).

AcknowledgementsThis work was supported in part by the Grant-in-Aid for Scientific Research on PriorityAreas ‘Nano Materials Science for Atomic-scale Modification 474’ from the Ministry ofEducation, Culture, Sports and Technology Japan (MEXT). N.S. acknowledges supportfrom PRESTO, Japan Science and Technology Agency, and the Industrial TechnologyResearch Grant Program in 2007 from the New Energy and Industrial TechnologyDevelopment Organization (NEDO), and the Grant-in-Aid for Young Scientists (A)(20686042) from MEXT. S.D.F. is supported as a Japan Society for the Promotionof Science (JSPS) fellow.

Author contributionsN.S. designed and carried out the STEM experiments and wrote the paper. S.D.F.carried out image simulations, image processing and wrote the paper. S.A. fabricatedthe bicrystal. S.A., T.M. and T.Y. supported the experiments and carried out densityfunctional theory calculations. Y.I. directed the entire study.

Additional informationSupplementary information accompanies this paper on www.nature.com/naturematerials.Reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions. Correspondence and requests for materials should beaddressed to N.S.

658 NATUREMATERIALS | VOL 8 | AUGUST 2009 | www.nature.com/naturematerials

© 2009 Macmillan Publishers Limited. All rights reserved.