cs-corrected stem investigation of dislocation core

Cs-Corrected Scanning Transmission Electron MicroscopyInvestigation of Dislocation Core Configurations at aSrTiO3/MgO Heterogeneous Interface

Yuanyuan Zhu,1 Chengyu Song,2 Andrew M. Minor,2,3 and Haiyan Wang,1,4,*

1Program of Materials Science and Engineering, Texas A&M University, College Station, TX 77843-3128, USA2National Center for Electron Microscopy, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA3Department of Materials Science and Engineering, University of California, Berkeley, CA 94720, USA4Department of Electrical and Computer Engineering, Texas A&M University, College Station, TX 7843-3128, USA

Abstract: Heterostructures and interfacial defects in a 40-nm-thick SrTiO3 ~STO! film grown epitaxially on asingle-crystal MgO ~001! were investigated using aberration-corrected scanning transmission electron micros-copy and geometric phase analysis. The interface of STO/MgO was found to be of the typical domain-matchingepitaxy with a misfit dislocation network having a Burgers vector of 1

2_ aSTO ^100&. Our studies also revealed that

the misfit dislocation cores at the heterogeneous interface display various local cation arrangements in terms ofthe combination of the extra-half inserting plane and the initial film plane. The type of the inserting plane,either the SrO or the TiO2 plane, alters with actual interfacial conditions. Contrary to previous theoreticalcalculations, the starting film planes were found to be dominated by the SrO layer, i.e., a SrO/MgO interface. Incertain regions, the starting film planes change to the TiO2/MgO interface because of atomic steps at the MgOsubstrate surface. In particular, four basic misfit dislocation core configurations of the STO/MgO system havebeen identified and discussed in relation to the substrate surface terraces and possible interdiffusion. Theinterface structure of the system in reverse—MgO/STO—is also studied and presented for comparison.

Key words: aberration-corrected scanning transmission electron microscopy ~Cs-corrected STEM!, geometricphase analysis ~GPA!, interface structure, dislocation core, perovskite, thin films

INTRODUCTION

The exact nature of the interface structure and its associateddefects has been a long-standing question for studies of thinfilm mechanical and physical properties ~Hirth & Pond,1996; Balzar et al., 2004!. Significantly different from theirbulk counterparts, many oxide thin films, almost all withheterogeneous interfaces, behave as strongly correlated elec-tronic systems ~Dagotto, 2005!. In these confined structures,multiple coupling states—charge ~electronic!, orbital ~and/orspin!, and inhomogeneous lattice strain—interact and be-come very sensitive to subtle structural perturbations andhence present novel properties ~Reiner et al., 2009!. Studieson typical perovskite SrTiO3 ~STO! single crystals haveshown that dislocation cores have a distinct local composi-tion ~Jia et al., 2005! and can switch the film electricalproperties ~Szot et al., 2006!. Ferroelectric barium stron-tium titanate ~Sr1�xBaxTiO3! grown on MgO ~Li et al.,2001! is considered as one of the most promising high-kdielectric candidates for replacing SiO2 in metal-oxide semi-conductor gates downsizing ~Escamilla et al., 2002!. Thefilm quality and functionality are largely dependent uponthe heterogeneous interface structure, i.e., interfacial defectsand lattice strains ~Szot et al., 2006; Schlom et al., 2007!.

Although rock salt MgO has one neutral ~001! surface,perovskite STO allows for two types of ~001! termination,

the SrO or the TiO2 planes, which could enable differentchemical bonding at the interface. For the interface contactof STO/MgO, previous first-principle calculations sug-gested that starting the STO film growth with a TiO2 layer ispreferred due to its thermodynamic stability ~Casek et al.,2004! and that it is in favor of strain relaxation ~Chenget al., 2002!, whereas experimentally it was not conclusive~Ernst, 1995!. In fact, due to the coupling effects at theconfined heterogeneous interface ~Reiner et al., 2009!, to-gether with the film-substrate lattice thermal expansiondifference and surface step terraces ~Dodson, 1988; Kienzleet al., 2009!, interfacial structures are typically complicatedand difficult to predict ~Chu et al., 2004!.

Owing to the development of aberration-corrected elec-tron microscopy, sub-angstrom resolution is readily achievedin high-resolution transmission electron microscopy~HRTEM! as well as in scanning transmission electronmicroscopy ~STEM! ~Erni et al., 2009!. Except for a fewspecial cases of channeling effect ~Fitting et al., 2006!, one ofthe most advantageous aspect of STEM is its capability offaithfully recording image contrast as a function of latticeatomic number Z, i.e., the contrast does not reverse for awide range of specimen thickness and imaging focus ~Kle-nov & Stemmer, 2006!. Therefore, it becomes possible todistinguish the atomic columns consisting of different ele-ments, such as the TiO and the Sr columns in STO. Due tothe large lattice mismatch between STO and MgO @about7.54% estimated by 2~as � af !/~as � af ! using bulk lattice

Received October 11, 2012; accepted February 8, 2013*Corresponding author. E-mail: [email protected]

Microsc. Microanal. 19, 706–715, 2013doi:10.1017/S1431927613000408 Microscopy AND

Microanalysis© MICROSCOPY SOCIETY OF AMERICA 2013

parameters at room temperature# , a misfit dislocation net-work is expected. Although the dislocation core structuresof STO were intensively studied in bulk samples ~Klie et al.,2009! or at homogeneous interfaces ~Zhang et al., 2002!,reports on misfit dislocations at the heterogeneous interfaceof functional oxides are limited to very few systems ~Miet al., 2007; Huijben et al., 2009!, and so far there is no suchreport for the interface of STO/MgO.

Here, we present a detailed interfacial structure analysison a pair of oxide heterostructures—STO/MgO and MgO/STO—using aberration-corrected high-resolution scanningtransmission electron microscopy ~Cs-corrected STEM! incombination with geometric phase analysis ~GPA! ~Hytchet al., 1998!. GPA is an image processing method used formapping lattice displacement and has been successfullyapplied to misfit dislocations and their associated strainfields ~Hytch et al., 2003; Sanchez et al., 2007; Dong & Zhao,2010!. Important details of the interface structure at theincoherent heterointerface, such as the termination planes,cation disorder, and nature of the substrate terraces, wereinvestigated at the atomic scale. In particular, we focused onthe misfit dislocations observed at the heterogeneous STO/MgO interface. The type of misfit dislocation network atthe heterointerface was determined by imaging in both the^100& and ^110& directions; and the local cation arrange-ments at the dislocation cores were further revealed byapplying GPA to Cs-corrected STEM images. Following thisthe correlations between the interface structure—the sub-strate steps and interdiffusion—and the misfit dislocationcore configurations were discussed and compared withatomic models. Microscopy analysis of the MgO/STO sys-tem is also presented as a unique comparison study forsystems with a reversed strain state.

MATERIALS AND METHODS

Two epitaxial thin films, STO on MgO and MgO on STO,were deposited by a standard pulsed-laser deposition tech-nique ~with a KrF excimer laser l � 248 nm! using bothcommercial MgO and STO targets ~Cerac, Inc., Milwaukee,WI, USA! and commercial single crystal MgO~100! andSTO~100! substrates. The depositions were carried out atthe same substrate-to-target distance of 5 cm, target tem-perature of 7808C, under an oxygen pressure of 27 Pa~about 200 mTorr!, with a laser energy of 300 mJ and arepetition rate of 2 Hz. Without post annealing, the sampleswere directly cooled down to room temperature in 4 �104 Pa ~300 Torr! of O2. After 720 s and 600 s depositions,resulting STO and MgO film thicknesses were around 40and 15 nm, respectively. Cross-sectional specimens wereprepared through a conventional TEM sample preparationroutine including cutting, gluing, grinding, polishing, andfinal precision ion polishing.

We employed the TEAM 0.5 TEM to investigate theinterfacial configuration. TEAM 0.5 is a modified FEI Titanmicroscope ~FEI Company, Hillsboro, OR, USA! with aspecial high-brightness Schottky-field emission electron

source and improved hexapole-type illumination aberra-tion corrector capable of resolving details as small as 0.05 nm~also the electron probe size! ~Kisielowski et al., 2008; Dah-men et al., 2009!. All annular dark-field ~ADF! STEM micro-graphs were recorded in TEAM 0.5 with a convergencesemi-angle of 17 mrad after fine-tuning of the probe correc-tor at 300 kV to a flat-phase angle of over 25 mrad @typicalresidual aberration coefficients are listed elsewhere ~Erniet al., 2009!# . A moderate ADF detector inner semi-angle of43 mrad was selected to boost the image signal-to-noiseratio for the later intensity line profile, while still maintain-ing a similar STEM intensity ratio as that at a higherdetector inner angle ~Klenov & Stemmer, 2006!.

To process the high-resolution ADF-STEM micro-graphs, GPA software ~a plug-in of Digital Micrograph 1.8.3package, HREM Research Inc., Higashimatsuyama, Japan!was employed to generate the corresponding phase andstrain maps. For each dot-like micrograph, the geometricphase algorithm reconstructs a displacement field utilizingFourier filtering centered around two noncollinear Braggvectors ~Hytch et al., 1998; Rouviere & Sarigiannidou, 2005!.The phase component of the displacement field, i.e., phaseimage, presents the local deviation in a chosen set of latticeplanes from a reference lattice; and the derivation of thelattice displacement field gives the strain field in a principaldirection. The “phase” in GPA does not refer to the electronwave function but to the position of image contrast max-ima. STEM imaging, due to its essential difference fromHRTEM projection imaging, tends to introduce artificialfringe distortions during image collection. The “flyback”error, which is along the fast scan direction, is one of themajor artifacts in scanning imaging ~Sanchez et al., 2006;Chung et al., 2010!. To minimize the “flyback” effects, wealigned the fast scan direction ~orthogonal to the slow scandirection! of each STEM image parallel to the principalx-axis of the strain map. This method was proven to beeffective in eliminating lattice distortion caused by “flyback”error in the STEM image with defect-related @�p, p# phaseboundaries ~Chung et al., 2010!.

RESULTS AND DISCUSSIONS

Figure 1a is a representative cross-sectional Cs-correctedSTEM image of a 38-nm-thick STO epitaxial thin film onMgO ~100! substrate acquired in ^100& zone axis. Since theSTEM image intensity is generally proportional to Z n of thescanned crystal ~n � 1.5 to 2.0!, the enlarged interfaceregion in Figure 1b shows the Sr columns as the brightestdots with the less bright TiO columns located at the centersof the Sr square lattice. Together Figure 1b presents thestandard perovskite lattice in ^100& projection. On the otherhand, the MgO columns have the least contrast, forminganother cubic lattice in the ^100& direction. ~Please note thatthe O-columns are not resolvable in the STEM images herebut illustrated in the atomic models in Fig. 1f.!

Five misfit dislocations, marked by red arrows in Fig-ure 1b, were found at the heterogeneous interface of STO/

Dislocation Core Configurations at a SrTiO3/MgO Heterogeneous Interface 707

MgO. To determine the dislocation network type, which isconsidered essential for describing the interface crystallo-graphic structure ~Cazottes et al., 2010!, GPA was appliedon the STEM images taken in both ^100& and ^110& direc-tions. In Figures 2a and 2b, typical Cs-corrected STEMimages of the STO/MgO heterointerface are presented, to-gether with corresponding Fourier transforms, filtered Braggimages ~Figs. 2c, 2d!, geometric phase images ~Figs. 2e, 2f!,and strain field maps ~Figs. 2g, 2h!. Compared with thefiltered Bragg images, the phase images provide a bettervisualization of the lattice variations from the reference~Hytch et al., 1998!, where in this case the crystalline MgOsubstrate was selected as a reference. As shown in the 020phase image ~Fig. 2e! of the STO/MgO interface in ^100&,the phase is approximately zero in the MgO substrate latticebut presents significant deviations in the top STO filmregion. These lattice displacements correspond to the misfitdislocations in the 020 Bragg image in Figure 2c. An obvi-

ous demonstration is also shown at the convergent regionof strain around the dislocation core in the in-plane-lattice-strain map Exx ~Fig. 2g!. Thus, the distribution of thesemisfit dislocations as well as their spacing can be easilydetermined using this analysis technique.

Two criteria can be used to identify the type of disloca-tion network: ~1! the direction in which the dislocation corecould be directly observed and ~2! the dislocation spacingrelationship ~Ernst, 1995!. Here, in Figures 2e and 2g, theperiodic edge dislocation cores were found in the ^100&direction, and their experimental spacing measured fromthe phase image was 2.67 nm, agreeing well with thetheoretical estimation @the misfit dislocation spacing ds �~af � as!/2~af � as! � 2.69 nm based on the bulk latticeparameters of the STO film and the MgO substrate# . Addi-tionally, in the ^110& direction, inspecting the STEM imagesby viewing the interface ~ N111! plane displacement variationin Figures 2d and 2f ~Chung et al., 2010! periodic disloca-

Figure 1. ~a! A representative high-resolution Cs-corrected STEM image of the STO/MgO heterointerface in ^100& zoneaxis. ~b! Enlarged interfacial area with five misfit dislocation cores. Red vertical arrows point to the extra-half planes ateach dislocation core. The corresponding matching domains were bracketed with their sizes noted by the number of theSTO ~020! and MgO ~020! lattice planes, respectively. A net lattice distortion angle of 0.78 and 0.38 was measured in theSTO layer and the MgO layer, respectively, about 5 nm away from the interface. ~c! The 020 phase image of the interfaceregion obtained by applying geometric phase analysis. ~d! Enlarged misfit dislocation from image b superimposed withthe corresponding phase image to show the relationship between the abrupt phase alternation and the position of theextra-half-plane. ~e! A Burgers circuit applied on the same dislocation core to illustrate the misfit dislocation has aBurgers vector of 1

2_ aSTO^010&. ~f! Schematic illustration of the dislocation core structure in image e by atomic models of

STO and MgO.

708 Yuanyuan Zhu et al.

tions are also visible. The measured spacing between these458 inclined dislocation end-on yields 3.77 nm ~similarlythe distance between the two components in Exx strain mapFig. 2g measured as 1.89 nm!, equal to the ds of ^100& afterprojecting in the ^110& direction. This confirms that thenature of the misfit dislocations at the STO/MgO hetero-interface is a ^100& dislocation network. More precisely, asshown in Figure 1b, there are 14 ~020!STO lattices that matchwith 13 ~020!MgO lattices cube-on-cube. This feature ofdomain-matching epitaxy ~DME! ~Narayan & Larson, 2003!leads to an estimated residual strain of 0.134%, which ismuch smaller than the rigid lattice-matching epitaxy strainof 7.836%.

In addition, the Cs-corrected high-resolution STEMimaging also reveals interesting chemical information includ-ing the type of the extra-half inserting plane for the misfitdislocations, which could determine the cation sublattice ofthe misfit dislocation core. The GPA technique has shownthe capability for illustrating the lattice variation reproduc-ibly ~Chung et al., 2010! and accurately ~Hytch et al., 2003!without human errors. When applying GPA to STEM im-ages, not only does the lattice displacement become morevisible, but the chemical characteristic of the lattice can alsobe presented simultaneously. For example, in Figure 1d, anenlarged image of the first dislocation core in Figure 1b issuperimposed by its 020 phase image ~Fig. 1d!, and thephase changes abruptly across an edge dislocation. This p

to �p ~purple to red! phase jump corresponding to atranslation of half a-lattice fringe spacing ~Hytch et al.,1998! is caused by inserting the extra-half-plane in the STOfilm. Moreover, here, based on the image intensity, it is aSrO extra-half-plane sitting on the dislocation core with aBurgers vector of 1

2_ aSTO^010& as demonstrated by the Burg-

ers circuit in Figure 1e. The atomic configuration of thisdislocation core is schematically illustrated in Figure 1f. Thesame methodology was applied to the Cs-corrected STEMimages in this work to unambiguously determine the extra-half-plane type. At the STO/MgO heterointerface, the insert-ing planes are identified as either the SrO extra-half-planeor the TiO2 extra-half-plane ~here, for example in Fig. 1b,three of SrO extra-half planes and two of TiO2 extra-half-planes!. Under the ideal case, in the absence of any pertur-bations, one single regular type of the extra-half-insertingplane, i.e., SrO or TiO2 plane, is expected at the dislocationcore centers. The types of inserting planes observed in thisstudy are possibly related to the interface imperfections,which is discussed later in the article.

Unlike the STO homogeneous dislocation core struc-tures ~Zhang et al., 2002; Klie et al., 2009!, the verticalinserting planes above the STO/MgO interface are observedalmost perfectly parallel to each other along the @010#direction. However, the STO ~002! stacking planes above theinterfacial dislocations are not perfectly horizontal but tilt-ing upward to the right, as shown by a curved open arrow

Figure 2. Typical Cs-corrected STEM images of the STO/MgO heterointerface in ~a! ^100& and ~b! ^110& zone axis. Theinserts are corresponding Fourier transforms. ~c! 020 Bragg image, ~e! 020 geometric phase image, and ~g! the in-planeExx strain field map calculated using both 020 and 002 vectors in image a. ~d! N111 Bragg image, ~e! N111 geometric phaseimage, and ~g! the in-plane Exx strain field map calculated using both N111 and 1 N11 vectors in image b.


in Figure 1b. This results in a net lattice tilting of less than18 depending on the imaging areas ~e.g., 0.78 for the area inFig. 1b!. Similar lattice bending was also observed in theMgO-substrate planes, with a smaller net tilting anglethrough the same region, reflecting a rumpled lattice distor-tion similar to a previous observation ~Jia et al., 2009!. Thiscan be attributed to the fact that the heterointerface isessentially a three-dimensional structure accommodatingresidual lattice strains in all a, b, and c directions; so that therumpled ~002! plane could be a result of the lattice displace-ment variation mainly in the out-of-plane c direction.

In short, the STO/MgO heterostructure with a latticemisfit larger than 7% ~at both room temperature and thedeposition temperature! is a typical DME interface, wherea misfit dislocations network with a Burgers vector of12_ aSTO ^100& is generated. This introduces possible structuraland chemical perturbations to the inhomogeneous strainfields.

Another key interfacial chemical feature at the STO/MgO heterogeneous interface is the starting film plane. Animage intensity profile along the atomic planes at the inter-face was performed on all of the acquired STO/MgO STEMimages. For example, two intensity line profiles at the inter-face are shown in Figures 3 and 4, respectively. According to

the intensity as well as the periodicity of each atomic plane~Figs. 3b, 4a!, the starting thin film planes in our observa-tions were dominated by the SrO plane, indicating a SrO/MgO contact at the interface. A previous study of the STOthin films with and without an atomic TiO2 buffer layer onMgO showed that the misfit dislocation spacings are approx-imately 3.1 nm and 2.8 nm, respectively ~McMitchell et al.,2009!. The latter case is closer to our observation of disloca-tion spacing, which confirms that the initial growth plane ofour STO film is dominated by SrO plane. However, ourexperimental observation is in contrast to some theoreticalcalculations, based on which the TiO2/MgO contact waspreferred for thermodynamic stability ~Casek et al., 2004!and for strain relaxation ~Cheng et al., 2002!. This inconsis-tency suggests that in real film growth, especially for thosesystems with comparable interface energy such as the SrO/MgO ~4.6 J/m2! and the TiO2/MgO ~3.1 J/m2! ~Sayle &Watson, 2001!, other factors such as interfacial defects thatare usually not taken into account in theoretical calculationscan influence the growth mechanism and the resultinginterface structure. Previous studies have already pointedout that by tuning the interfacial chemistry it is possible tomodify the strain distribution through relaxation ~Kim et al.,1999!.

Figure 3. ~a! Intensity line profiles along the six atomic planes noted as 1 to 6 at ~b! the interface of STO/MgO inFigure 1b. The MgO substrate surface terrace with atomic-height steps is marked by a dotted line to guide the eye. Theextra-half planes identified by GPA are marked by arrows; solid arrows represent SrO-planes and broken arrows are forTiO2-planes. ~c! Enlarged area from image b shows three different dislocation core configurations. ~d! The fourth typeof core configuration for the STO/MgO heterostructures from a different region of the specimen. Atomic-column-projection models of the four basic dislocation core configurations of the STO/MgO system are illustrated schematicallyto show their composition and core size.


We explored the above issue by a careful study on themajor interfacial defects at the heterointerface STO/MgOand found that the configuration of the dislocation coresis closely related to the local interfacial conditions, i.e.,substrate surface terraces and possible substrate/film inter-diffusion. In Figure 3b, it was found that as steps withatomic-scale height emerge at the MgO-substrate surface,instead of maintaining the same type of starting plane andgenerating antiphase boundaries ~Mi et al., 2007!, the inter-face contact changes from SrO/MgO to TiO2/MgO, whichalso leads to a different core configuration. It should benoted that for both Figures 3 and 4, the color-coded circlesin the core models represent the dominant atom type ineach atomic column; there could be intermixing along theincident e-beam direction. In fact, additional intensities inthe starting SrO-layer ~marked by stars in Figs. 3 and 4! andstronger intensities in the MgO-layers ~arrows in Fig. 4!suggest possible interfacial interactions. A typical case ofthis interface interdiffusion is shown in Figure 5. The ob-served interdiffusion, up or down, at the interface suggeststhat it is more likely to happen at regions slightly away fromthe dislocation cores. Therefore, four basic misfit disloca-tion core configurations—SrO-~SrO/MgO!, TiO2-~TiO2/

MgO!, SrO-~TiO2/MgO!, and TiO2-~SrO/MgO!—dependingon the interface types as well as the extra-half insertingplanes, were observed in the specimen ~Figs. 3c, 3d!. Todistinguish the two, different notations are adopted, “/” forinterface and “-” for the extra-half inserting plane. Forexample, in Figure 3c, SrO-~SrO/MgO! stands for a disloca-tion core that is formed by inserting the SrO-plane on topof the heterogeneous interface contact of SrO/MgO. Sincethe interface contact of SrO/MgO was dominant in theSTO/MgO system, the two core configurations of SrO-~SrO/

Figure 4. ~a! The STEM image of a different region of the STO/MgO interface. The extra-half planes identified by GPA~not shown here! are marked by arrows. ~b! The intensity line profiles of image a. Additional intensities in the startingSrO plane and the surface MgO plane marked by stars and arrows, respectively, suggest a certain extent of interfacialinterdiffusion. ~c! Enlarged dislocation core area from image b. ~d! An atomic-projection model illustrates theinterdiffusion observed in image c.

Figure 5. A typical STO/MgO interface region showing STO dif-fusion down into the substrate layers. The dislocation core posi-tions are circled by dotted lines, and the diffusion of STO into thesubstrate is noted by open arrows.


MgO! and TiO2-~SrO/MgO!, with no preference betweenthe two, are observed more often among all 20 imagesanalyzed. Furthermore, it is interesting to see that the dislo-cation cores with the same type of initial film layer and ofthe extra-half inserting plane present a smaller core [email protected]., the three-atom-consisted core structure of SrO-~SrO/MgO! and TiO2-~TiO2/MgO!# compared to the ones thatconsisted of different planes @i.e., the SrO-~TiO2/MgO! andTiO2-~SrO/MgO! cores consisted of five atoms# . This sug-gests a direct connection between the chemical disorder andlattice displacement at the dislocation core, i.e., a greaterlattice distortion happens when more heterogeneous com-ponents are confined at the interface.

The above observed correlations between the real sub-strate surface conditions with the misfit dislocation positionand core configuration are illustrated schematically by atomicmodels of the STO/MgO heterogeneous interface in Fig-ure 6. Under the ideal condition ~Fig. 6a!, for the observeddominant interface contact of SrO/MgO in this study, thespecific 14~020!STO/13~020!MgO domain matching is sup-posed to allow only one type of the dislocation core configu-ration, i.e., either the SrO-~SrO/MgO! or the TiO2-~SrO/MgO!. However, in reality the substrate surface is notperfectly smooth. Based on the observations in Figure 3c, acorresponding illustration is shown in Figure 6b that shows

that the emergence of atomic steps at the substrate surfacecould alternate the interface contact, and thus the core con-figuration changes accordingly. In addition, interdiffusioncan also happen at the substrate/film interface ~e.g., Fig. 5!.Interdiffusion brings additional compositional disorders tothe interface chemistry, disturbing the ideal dislocation coreperiodicity ~Fig. 6c!, as well as the dislocation spacing. Insummary, a certain amount of interdiffusion and atomic-height substrate steps were observed at the STO/MgO hetero-geneous interface with a high density of misfit dislocations.Despite the interfacial defects and rich dislocation core con-figurations, the absence of lattice defects such as antiphaseboundaries or pores found in other pulsed labor deposition~PLD! grown STO/MgO systems ~Tse et al., 2010! indicatesthe high quality of the STO epitaxial thin films observed inthis study. It was suggested that a monolayer of TiO2 needsto be deposited on the MgO ~100! substrate prior to STO inorder to achieve a two-dimensional film growth ~McMitch-ell et al., 2009!. However, under the PLD conditions used inthis study, there did not seem to be any problem for growinghigh quality STO film on MgO~100! with thicknesses rang-ing from 38 nm to over 150 nm.

For comparison, the reversed heterogeneous systemMgO/STO was also studied using the STEM and GPAtechnique. A representative high-resolution Cs-correctedSTEM image of the MgO/STO~100! and the corresponding020 Bragg image, phase image and Exx strain map of theinterface segment are shown in Figures 7a–7d. This 15-nm-thick MgO thin film deposited on a single crystal STO~001!substrate also shows high quality epitaxial growth and anatomically sharp interface. Similar to the former STO/MgOsystem, the average spacing between the misfit dislocationsclearly located by the abrupt contrast in the 020 phaseimage ~Fig. 7c! is about 2.70 nm, suggesting the STO/MgOalso has a semicoherent DME interface with a ^100& misfitdislocation network. It is notable that the phase changeacross the STO/MgO dislocation is from �p to p, oppositeto the direction in the previous case of MgO/STO ~Fig. 1c!,and corresponds to the “missing” ~020! planes in the MgOthin film as enlarged in Figure 7b. In Figure 8a, the extra-half planes in the STO substrate were marked by arrows toindicate the misfit dislocation cores. These misfit disloca-tions are generated to accommodate the lattice mismatchfrom the MgO film above. As in the reversed system, thetype of the extra-half planes observed here still show amixture of the SrO and TiO2 planes. As for the interfacecontact, the image intensity profiles along the atomic planesat this interface in Figure 8b suggest a MgO/TiO2 interface,which may be largely due to the fact that the terminationlayer of the STO~001! substrate is dominated by the TiO2-plane, and there is little surface modification prior to thedeposition of the MgO film.

CONCLUSIONS

In summary, we used high-resolution Cs-corrected STEMimaging and GPA to study the interface structure of two

Figure 6. ~a! The ideal condition of the 14~020!STO/13~020!MgOdomain matching on a perfectly flat interface. Only one type ofdislocation core configuration is expected. ~b! The case of theMgO-substrate surface has an atomic-height step. The change ofinterface contact type alters core configuration. ~c! The case ofinterdiffusion and its effect on dislocation core configuration.


STO and MgO heterosystems. Both thin films show excel-lent 14STO~020!/13MgO~020! DME with a high density^100& misfit dislocation network ~dislocation spacing around2.67 nm!. The high-resolution STEM micrographs at theheterogeneous interfaces suggest that, in both cases, thetype of extra-half planes at the dislocation core centerstrongly depends on the actual interfacial microstructures.The contact interface is dominated by SrO/MgO in theSTO/MgO system and by MgO/TiO2 in the case of theMgO grown on STO. Further analysis of the dislocationcore configuration of the STO/MgO interface revealed a

highly sensitive interfacial environment, where minor geo-metrical ~surface atomic-height steps! or compositional ~lo-calized interdiffusion! and/or charge perturbations couldaffect the interface contact and therefore the dislocationcore structure. The characterization method used in thiswork, Cs-corrected STEM combined with GPA processing,can be applied in the study of nanostructures involvingstrain and stoichiometry modifications besides perovskitesystems. The experimental observations provide actual de-fect structures in atomic scale that could be used for theo-retical computation work.

Figure 7. ~a! Typical Cs-corrected STEM images of the MgO/STO~001! heterointerface along ^100& zone axis. ~b! Corre-sponding 020 Bragg image, ~e! 020 geometric phase image, and ~g! the in-plane Exx strain field map calculated using both020 and 002 vectors. One misfit dislocation is enlarged to show the “missing” ~020! planes in the MgO thin film.

Figure 8. ~a! The STEM image of a different region of the MgO/STO~001! interface. The extra-half planes identified byGPA ~not shown here! are marked by arrows; solid arrows represent SrO planes and broken arrows are for TiO2-planes.~b! The intensity line profiles along the atomic planes in image a. It suggests an MgO/TiO2 contact.


ACKNOWLEDGMENTS

This research was funded by National Science Foundation~Ceramic Program Award No. NSF-0846504! and the AirForce Office of Scientific Research ~Contract No.: FA9550-09-1-0114!. The microscopy experiments were performed atthe National Center for Electron Microscopy ~NCEM!, whichis supported by the Office of Science, Office of Basic EnergySciences of the U.S. Department of Energy under ContractNo. DE-AC02-05CH11231. Y.Z. is grateful to C. Kisielowskiat NCEM for helpful guidance and discussions on thiswork.

REFERENCESBalzar, D., Ramakrishnan, P.A. & Hermann, A.M. ~2004!. Defect-

related lattice strain and the transition temperature in ferroelec-tric thin films. Phys Rev B 70, 0921031–0921034.

Casek, P., Bouette-Russo, S., Finocchi, F. & Noguera, C. ~2004!.SrTiO3~001! thin films on MgO~001!: A theoretical study. PhysRev B 69, 0854111.

Cazottes, S., Zhang, Z.L., Daniel, R., Chawla, J.S., Gall, D. &Dehm, G. ~2010!. Structural characterization of a Cu/MgO~001!interface using C-S-corrected HRTEM. Thin Solid Films 519,1662–1667.

Cheng, C., Kunc, K., Kresse, G. & Hafner, J. ~2002!. SrTiO3/MgO~001! and MgO/SrTiO3~001! systems: Energetics andstresses. Phys Rev B 66, 0854191–0854198.

Chu, M.W., Szafraniak, I., Scholz, R., Harnagea, C., Hesse, D.,Alexe, M. & Gosele, U. ~2004!. Impact of misfit dislocationson the polarization instability of epitaxial nanostructured fer-roelectric perovskites. Nat Mater 3, 87–90.

Chung, J.H., Lian, G.D. & Rabenberg, L. ~2010!. Practical andreproducible mapping of strains in Si devices using geometricphase analysis of annular dark-field images from scanningtransmission electron microscopy. IEEE Electron Device Lett 31,854–856.

Dagotto, E. ~2005!. Complexity in strongly correlated electronicsystems. Science 309, 257–262.

Dahmen, U., Erni, R., Radmilovic, V., Kisielowski, C., Rossell,M.D. & Denes, P. ~2009!. Background, status and future of thetransmission electron aberration-corrected microscope project.Philos Trans R Soc A 367, 3795–3808.

Dodson, B.W. ~1988!. Nature of misfit dislocation sources instrained-layer semiconductor structures. Appl Phys Lett 53,394–396.

Dong, Z.S. & Zhao, C.W. ~2010!. Measurement of strain fields inan edge dislocation. Physica B 405, 171–174.

Erni, R., Rossell, M.D., Kisielowski, C. & Dahmen, U. ~2009!.Atomic-resolution imaging with a sub-50-pm electron probe.Phys Rev Lett 102, 0961011–0961014.

Ernst, F. ~1995!. Metal-oxide interfaces. Mater Sci Eng R-Rep 14,97–156.

Escamilla, R., Akachi, T., Gomez, R., Marquina, V., Mar-quina, M.L. & Ridaura, R. ~2002!. Suppression of T-c in the~Y0.9Ca0.1!Ba2Cu4�xFexO8 system. Supercond Sci Technol 15,1074–1080.

Fitting, L., Thiel, S., Schmehl, A., Mannhart, J. & Muller,D.A. ~2006!. Subtleties in ADF imaging and spatially resolvedEELS: A case study of low-angle twist boundaries in SrTiO3.Ultramicroscopy 106, 1053–1061.

Hirth, J.P. & Pond, R.C. ~1996!. Steps, dislocations and disconnec-tions as interface defects relating to structure and phase trans-formations. Acta Mater 44, 4749–4763.

Huijben, M., Brinkman, A., Koster, G., Rijnders, G., Hilgen-kamp, H. & Blank, D.H.A. ~2009!. Structure-property relationof SrTiO3/LaAlO3 interfaces. Adv Mater 21, 1665–1677.

Hytch, M.J., Putaux, J.L. & Penisson, J.M. ~2003!. Measurementof the displacement field of dislocations to 0.03 angstrom byelectron microscopy. Nature 423, 270–273.

Hytch, M.J., Snoeck, E. & Kilaas, R. ~1998!. Quantitative mea-surement of displacement and strain fields from HREM micro-graphs. Ultramicroscopy 74, 131–146.

Jia, C.L., Mi, S.B., Urban, K., Vrejoiu, I., Alexe, M. & Hesse, D.~2009!. Effect of a single dislocation in a heterostructure layeron the local polarization of a ferroelectric layer. Phys Rev Lett102, 1176011–1176014.

Jia, C.L., Thust, A. & Urban, K. ~2005!. Atomic-scale analysis ofthe oxygen configuration at a SrTiO3 dislocation core. Phys RevLett 95, 2255061–2255064.

Kienzle, D., Quezada, B. & Marks, L. ~2009!. SrTiO3 ~001! ~root13 � root 13! R33.7 degrees surface reconstruction. MicroscMicroanal 15, 1022–1023.

Kim, D.W., Kim, D.H., Kang, B.S., Noh, T.W., Lee, D.R. & Lee,K.B. ~1999!. Roles of the first atomic layers in growth of SrTiO3

films on LaAlO3 substrates. Appl Phys Lett 74, 2176–2178.Kisielowski, C., Freitag, B., Bischoff, M., van Lin, H., Lazar,

S., Knippels, G., Tiemeijer, P., van der Stam, M., von Har-rach, S., Stekelenburg, M., Haider, M., Uhlemann, S.,Muller, H., Hartel, P., Kabius, B., Miller, D., Petrov, I.,Olson, E.A., Donchev, T., Kenik, E.A., Lupini, A.R., Bentley,J., Pennycook, S.J., Anderson, I.M., Minor, A.M., Schmid,A.K., Duden, T., Radmilovic, V., Ramasse, Q.M., Watanabe,M., Erni, R., Stach, E.A., Denes, P. & Dahmen, U. ~2008!.Detection of single atoms and buried defects in three dimen-sions by aberration-corrected electron microscope with 0.5-angstrom information limit. Microsc Microanal 14, 469–477.

Klenov, D.O. & Stemmer, S. ~2006!. Contributions to the contrastin experimental high-angle annular dark-field images. Ultra-microscopy 106, 889–901.

Klie, R.F., Walkosz, W., Yang, G. & Zhao, Y. ~2009!. Aberration-corrected Z-contrast imaging of SrTiO3 dislocation cores. JElectron Microsc 58, 185–191.

Li, H., Roytburd, A.L., Alpay, S.P., Tran, T.D., Salamanca-Riba,L. & Ramesh, R. ~2001!. Dependence of dielectric propertieson internal stresses in epitaxial barium strontium titanate thinfilms. Appl Phys Lett 78, 2354–2356.

McMitchell, S.R.C., Tse, Y.Y., Bouyanfif, H., Jackson, T.J.,Jones, I.P. & Lancaster, M.J. ~2009!. Two-dimensional growthof SrTiO3 thin films on ~001! MgO substrates using pulsedlaser deposition and reflection high energy electron diffraction.Appl Phys Lett 95, 1741021–1741023.

Mi, S.B., Jia, C.L., Faley, M.I., Poppe, U. & Urban, K. ~2007!.High-resolution electron microscopy of microstructure ofSrTiO3/BaZrO3 bilayer thin films on MgO substrates. J Crystal-logr Growth 300, 478–482.

Narayan, J. & Larson, B.C. ~2003!. Domain epitaxy: A unifiedparadigm for thin film growth. J Appl Phys 93, 278–285.

Reiner, J.W., Walker, F.J. & Ahn, C.H. ~2009!. Materials scienceatomically engineered oxide interfaces. Science 323, 1018–1019.

Rouviere, J.L. & Sarigiannidou, E. ~2005!. Theoretical discus-sions on the geometrical phase analysis. Ultramicroscopy 106,1–17.


Sanchez, A.M., Galindo, P.L., Kret, S., Falke, M., Beanland, R.& Goodhew, P.J. ~2006!. An approach to the systematic distor-tion correction in aberration-corrected HAADF images. JMicrosc-Oxf 221, 1–7.

Sanchez, A.M., Lozano, J.G., Garcia, R., Herrera, M., Ruffen-ach, S., Briot, O. & Gonzalez, D. ~2007!. Strain mapping atthe atomic scale in highly mismatched heterointerfaces. AdvFunct Mater 17, 2588–2593.

Sayle, D.C. & Watson, G.W. ~2001!. The atomistic structures ofMgO/SrTiO3~001! and BaO/SrTiO3~001! using simulated amor-phization and recrystallization. J Phys Chem B 105, 5506–5514.

Schlom, D.G., Chen, L.Q., Eom, C.B., Rabe, K.M., Streiffer, S.K.& Triscone, J.M. ~2007!. Strain tuning of ferroelectric thinfilms. Ann Rev Mater Res 37, 589–626.

Szot, K., Speier, W., Bihlmayer, G. & Waser, R. ~2006!. Switch-ing the electrical resistance of individual dislocations in single-crystalline SrTiO3. Nat Mater 5, 312–320.

Tse, Y.Y., McMitchell, S.R.C., Jackson, T.J., Jones, I.P. & Genc,A. ~2010!. Microstructural investigation of strontium titanatefilms grown by interval pulsed laser deposition. In ElectronMicroscopy and Analysis Group Conference 2009, Baker, R.T.~Ed.!, pp. 0120401–0120404.

Zhang, Z.L., Sigle, W. & Ruhle, M. ~2002!. Atomic and elec-tronic characterization of the a 100 dislocation core in SrTiO3.Phys Rev B 66, 0941081–0941088.


cs-corrected stem investigation of dislocation core

Documents