14. HIGH RESOLUTION TRANSMISSION ELECTRON MICROSCOPYmx.nthu.edu.tw/~zhwei/member/handbook/Ch14_427-453.pdf · 2008-07-07 · 14. High Resolution Transmission Electron Microscopy 429

14. HIGH RESOLUTION TRANSMISSION ELECTRON MICROSCOPY

DAVID J. SMITH

1. HRTEM AND NANOTECHNOLOGY

With the rapidly escalating attention recently being given to nanoscale scienceand technology, techniques capable of structural characterization on the nanometerscale have central importance. The high-resolution transmission electron microscope(HRTEM) has evolved over many years to such an extent that resolving powers ator close to the one Ångstrom (0.1 nm) level can nowadays be attained almost on aroutine basis. Using correct operating conditions and well-prepared samples, high-resolution image characteristics are interpretable directly in terms of projections ofindividual atomic-column positions. With quantitative recording and suitable imageprocessing, atomic arrangements at defects and other inhomogeneities can be reli-ably and accurately determined. Since nanoscale irregularities have a marked influenceon bulk behavior, the HRTEM has become a powerful and indispensable tool forcharacterizing nanostructured materials. This chapter first outlines some of the basictheoretical principles and practical aspects of the HRTEM. Representative applicationsto problems involving nanostructured materials are then described. Ongoing trendsare identified and some underlying problems associated with use of the HRTEM arebriefly discussed. The interested reader is referred to review articles [1–4], conferenceproceedings [5–10] and research monographs [11–13] for further information aboutHRTEM and additional applications.

428 II. Electron Microscopy

2. PRINCIPLES AND PRACTICE OF HRTEM

2.1. Basis of Image Formation

The process of image formation in the HRTEM can be considered as occurringin two stages. Incoming or incident electrons undergo interactions with atoms ofthe specimen, involving both elastic and inelastic scattering processes. The electronwavefunction which leaves the exit surface of the specimen is then transmitted throughthe objective lens and subsequent magnifying lenses of the electron microscope to formthe final enlarged image. Our attention here will be focused on those electrons that areelastically scattered since these contribute to formation of the high-resolution bright-field image. Note, however, that the inelastically scattered electrons can also be usedto provide invaluable information about the sample composition using the techniqueof electron-energy-loss spectroscopy (EELS). Moreover, electrons scattered to verylarge angles are utilized for Z-contrast annular-dark-field imaging in the scanningtransmission electron microscope (STEM). These possibilities are explored elsewherein other chapters.

Electron scattering is a strongly dynamical process, unlike X-ray or neutron scatter-ing, so that a simple kinematical scattering approximation is insufficient for understand-ing image formation for all but the thinnest of samples. Multiple electron scattering andlarge phase changes are typical for most samples, which means that the relative heightsof the different atoms in the specimen must be taken into account when attemptingquantitative image interpretation. Image simulations, which must also take into accountadditional effects caused by the objective lens, are essential before detailed informationabout atomic arrangements at crystal defects such as dislocations and interfaces can beconfirmed. Several alternative approaches to image simulation have been developed[14, 15]. The most widespread is based on n-beam dynamical theory [16, 17], and istermed the multislice approach. In this theory, atoms in the specimen are consideredas located on narrowly-separated planes (or slices), normal to the beam direction. Theelectron wavefunction is then propagated slice-by-slice through the sample to eventu-ally form the exit-surface wavefunction. This iterative process lends itself to convenientcomputer algorithms which enable rapid computations to be carried out, which areespecially useful during the refinement of unknown defect structures [18, 19]. Furtherinformation about other equivalent approaches to electron scattering can be found inthe monograph by Cowley [20].

The electron wavefunction at the exit surface of the specimen must still be transferredto the final viewing screen or recording medium. This process is dominated by thetransfer characteristics of the objective lens of the microscope. The effect of this lenson image formation is conveniently understood by reference to what is termed thephase contrast transfer function (TF), as described by Hanszen [21]. The TF is bothspecimen- and microscope-independent so that a single set of universal curves enablesthe transfer characteristics of all objective lenses to be described. Electron microscopesthat have different objective lenses, or operate at different accelerating voltages, areeasily compared by using suitable scaling factors. Figure 1 shows TFs computed forthe optimum defocus of the objective lens for a typical 400-kiloVolt HRTEM. Two

14. High Resolution Transmission Electron Microscopy 429

Figure 1. Transfer functions for high-resolution phase-contrast imaging at optimum defocus for objectivelens (Cs = 1.0 mm) of 400-kV HREM: (a) coherent illumination; and (b) partially coherent illumination,with focal spread of 8 nm and incident beam convergence half-angle of 0.5 mrad. Arrow indicatesinterpretable resolution limit.

cases are shown, corresponding to: (a) coherent, and (b) partially coherent, incidentelectron illumination. Notice that the TF has an oscillatory nature, meaning thatelectrons scattered to different angles undergo relative phase reversals, which wouldlead directly to artefactual detail in the final image. Note also that the shape of the TF isfurther affected by defocus changes, which result in additional phase changes that willalter the image appearance. It is thus essential to have an accurate knowledge of the lensdefocus, since much of the detail contained in the recorded high-resolution image willotherwise be uninterpretable. In the ideal case, the incident electron beam will be acoherent, monochromatic plane wave. In practice, some inevitable loss of coherencewill result from focal spread (temporal coherence) and finite beam divergence (spatialcoherence) [22, 23]. Mathematically, the effects of partial coherence can be representedby envelope functions. These will cause dampening of the TF for spatial frequenciescorresponding to larger scattering angles, as illustrated, for example, by curve (b) inFig. 1. Note also that the positions of the TF crossovers, or “zeroes”, are not affectedby the envelope functions. However, specimen information that is scattered to higherspatial frequencies, which would be equivalent to higher image resolution, is liable tobe lost during image formation. Lanthanum hexaboride has been used as the electronsource for calculation of the curve (b) with partial coherence. As discussed in thefollowing section, additional specimen information can possibly become available byusing a high-coherence electron source such as the field-emission gun (FEG).

2.2. Definitions of Resolution

Resolution is traditionally discussed in terms of the ability of an imaging system to dis-criminate between two discrete objects, and is usually closely coupled to the wavelengthof the incident illumination. Since high-energy electrons have picometer wavelengths,resolution limits on the same scale might be anticipated. In practice, electron lenses


have unavoidable aberrations, and perfectly coherent electron sources are unattainable.It then becomes necessary to make a compromise between the traditional diffractionlimit, which varies inversely as the aperture angle, and spherical aberration, whichvaries rapidly with the cube power of the angle. The end result is an expression of theform

d = ACs1/4λ3/4 (1)

where Cs is the spherical aberration coefficient of the objective lens, λ is the electronwavelength, and the value of the constant A depends on certain assumptions that aremade about the imaging conditions.

In practice, there are several resolution limits that are applicable to any specificHRTEM [24]. These are conveniently understood by reference to the TF of theobjective lens. The interpretable resolution, which is sometimes known as the structuralor point resolution, is defined by the position of the first zero crossover of the TF atthe optimum defocus. This resolution is indicated by the arrow in Fig. 1. This specificdefocus gives the widest possible band of spatial frequencies without a phase reversal,and the corresponding interpretable resolution is then given by δ ∼ 0.66(Cs λ

3)1/4.Values of Cs inevitably increase as the electron energy is increased due to limits onthe saturation of the pole-piece material. However, overall improvements in δ shouldbe obtained due to reductions in λ as the accelerating voltage is increased. Typicalresolution figures using realistic Cs values (which range from 0.3 mm at 100 keV toabout 1.5 mm at 1.0 MeV) are in the range of 0.25 to 0.12 nm. Other practical factorssuch as the size and cost of higher-voltage electron microscopes, as well as the increasinglikelihood of electron irradiation damage due to the higher energy electron beam,become important considerations. Intermediate-voltage HRTEMs which operate inthe range of 200 to 400 kV have become widespread due to the impact of these factors.

The instrumental resolution or information limit of the HRTEM is usually defined interms of the damping produced by the envelope functions. A value of approximately15% [i.e., exp(-2)] is commonly taken as the resolution cutoff since this level is regardedas the minimum acceptable for image processing requirements [25]. In recent 200- or300-kV HRTEMs equipped with an FEG electron source, this resolution limit extendswell beyond the interpretable resolution. Because of the oscillatory nature of the TF, thevery fine detail present in the image at such resolution levels will, however, not be easilyrelated to specimen features. Higher-order diffracted beams with inverted phase canbe prevented from contributing to the image by using an objective aperture of suitablediameter located in the back focal plane of the objective lens. Conversely, providedthat the defocus and Cs values are known with sufficient accuracy, then the phase-modulating effects of the TF can be removed by a posteriori image processing. Imageinterpretability with improved resolution can be achieved, as demonstrated by thepioneering work of Coene et al. [26] who used a focal-series reconstruction approachto resolve oxygen atoms for the first time in a high-temperature superconductor.

The term lattice-fringe resolution refers to the finest spacings of the lattice fringes thatare visible in high-resolution images from a particular HRTEM. These lattice fringes


result from the interference between two or more diffracted beams from a crystallinematerial. This resolution is determined by the overall instrumental stability of theHRTEM, and is often a direct reflection of freedom of the microscope environmentfrom adverse external factors such as acoustic noise, mechanical vibrations, and straymagnetic fields. It must be noted, however, that lattice fringes with very fine spacingsdo not usually provide any useful information about local atomic arrangements. Theinterfering diffracted beams typically originate from comparatively large specimenareas, and the lattice fringe images may be recorded at significant underfocus conditionswhen there are many TF oscillations. For many years, the lattice-fringe resolution waswidely regarded as the ultimate figure of merit for an HRTEM. It should be appreciatedthat the interpretable and instrumental resolution limits are nowadays considered farmore useful for comparison purposes.

2.3. Lattice Imaging or Atomic Imaging

Most elemental and compound materials have unit cells that are relatively large incomparison with the resolution limits of commercially available HRTEMs. High-resolution lattice-fringe images are thus relatively straightforward to obtain when thesematerials are viewed in major low-index projections. However, it is important toappreciate that only a very small subset of these lattice images can be interpreted interms of atomic arrangements. The basic problem is that the phase reversals that resultfrom TF oscillations when the defocus of the objective lens is changed make it difficultto recognize or select the appropriate optimum defocus. The characteristic image of acrystal defect or the Fresnel fringe along the edge of the sample is often indispensablefor recognizing a particular defocus setting. Some prior knowledge or calibration of thefocal step size corresponding to the finest focus control is thus essential. The situationis even more complicated for materials with small unit cells which have comparativelyfew diffracted beams contributing to the final image. For these materials, identicalimages, referred to as Fourier or self-images, recur periodically with changes in defocus,with the period given by 2d2/λ, where d is the corresponding lattice spacing [27, 28].With a half-period change in defocus, all image features reverse in contrast, with blackspots becoming white and vice versa. As an example, a series of image simulations for[001] tin dioxide is shown in Fig. 2, where the major reversals in contrast of the imagefeatures from white to black to white to black are readily apparent [29].

Further misleading complications are likely to arise in thicker crystals because ofdynamical multiple scattering. As the sample thickness is increased, intensity will beprogressively lost from the directly transmitted beam, and it will instead build up inthe diffracted beams. Eventually, a thickness is reached where the direct beam will berelatively low in intensity (in the vicinity of what is termed the thickness extinctioncontour), and the resulting image will be dominated by (second-order) interferenceprocesses between the various diffracted beams. In some small-unit-cell materials, theseinterferences can often lead to pairs of white spots, graphically referred to as dumbbells,which appear to portray faithfully all atom positions in the unit cell [30]. Except underhighly specific thickness and defocus values, it has been found that the separation


Figure 2. Through-focal series of image simulations for crystal of tin dioxide in [100] projection:500 keV, Cs = 3.5 mm, crystal thickness −2.8 nm. Focus extends in 10 nm steps from −140 nm (top left)to +10 nm (bottom right) [29].

between these spots will invariably be slightly different from the correct projectedatomic separations for the specific material [31]. Moreover, erroneous image featurescan also be anticipated to occur at structural discontinuities such as interfaces [32].Such interference lattice-fringe images from thicker specimen regions should thus beconsidered as nothing more than interesting coincidental effects and they should neverever be interpreted in terms of atomic positions.

2.4. Instrumental Parameters

In common with other sophisticated pieces of equipment, the HRTEM has an arrayof adjustable parameters that must often seem almost overwhelming to the newcomer.However, only a small subset must be known with any degree of accuracy. Experimentaldetermination of these instrumental parameters is well documented elsewhere [24],so these details are not reproduced here. For atomic-resolution imaging, it is essen-tial that the accelerating voltage be fixed and highly stable, preferably to within onepart per million (ppm), but the exact value is relatively unimportant. A similar sit-uation holds for the current of the objective lens, which will determine its focallength and several other imaging parameters, including Cs , the spherical aberrationcoefficient. In practice, it is highly desirable to operate with a fixed objective lenscurrent, due to the extreme sensitivity of several adjustable parameters such as theincident-beam tilt alignment and the objective lens astigmatism to the exact currentsetting. The objective lens current should thus be monitored continuously, and whenthe sample is tilted or another field of view is selected, the sample height shouldbe adjusted rather than altering the objective lens current. Knowledge of the objectivelens defocus is critical to image interpretation because of the extreme sensitivity of theimage appearance to the defocus value. The highly characteristic image appearance


of complicated structures at the optimum defocus can be recognized in some specialcases, but for unknown aperiodic features such as a dislocation or grain boundary, someother means for selecting defocus is required. It has become increasingly common togenerate a through-thickness-through-focus tableau of images, at least of the perfectcrystal structure, before commencing microscopy [33]. Some electron microscopistshave recommended recording a focal series of images with pre-calibrated focal stepsizes, which thereby reduces somewhat the subjective element of the matching pro-cess [34]. Methods based on cross-correlation [35], and non-linear least squares [36]have been developed for matching experimental and simulated images, and these canprovide determination of both defocus and local specimen thickness on a local scale.

At extreme resolution limits, several higher-order objective-lens aberrations have adominant role in determining the overall image integrity. Two-fold image astigmatismis well-known to result in focal length differences in orthogonal image directions, witha corresponding effect on the image appearance, but its presence can usually be detectedby reference to the image of a thin amorphous material, such as carbon or germanium:Fourier transformation reveals a set of concentric rings, that are more or less ellipticaldepending on the amount of astigmatism present [37]. The successful implementationof image reconstruction schemes involving either focal series [26] or off-axis electronholography [38] requires knowledge of the spherical aberration coefficient with anerror of no greater than 1%. The graphical method of optical diffractogram analysis[37] is no longer sufficiently accurate, and alternative methods are required in order toreach the desired level of accuracy [39]. The third-order aberrations of axial coma andthree-fold astigmatism are more difficult to detect and quantify because their influenceis only clearly visible in images of amorphous materials recorded with tilted incidentillumination [40, 41]. Detection and correction of third-order aberrations is discussedfurther in Section 3.4.

2.5. Further Requirements

There are additional requirements that must be satisfied before the atomic-scaleresolution of the HRTEM can be fully utilized for characterizing nanostructured mate-rials. The sample region being imaged must be sufficiently thin, typically meaning thatits thickness should be on the order of 15 nm or less depending on its composition.Otherwise, multiple electron scattering effects become significant, and the very finedetails present in the image are not likely to be interpretable in terms of atomic arrange-ments, even with the assistance of image simulations. Since the final recorded imagerepresents a two-dimensional projection of the crystal structure, the incident electronbeam must be aligned closely with a major zone axis of the crystal. The tolerance foradjustment of the crystal alignment to avoid overlap of projected atomic columns obvi-ously becomes more demanding for thicker crystals and smaller column separations[42]. The examination of crystalline defects must be confined to planar faults, such astwin boundaries and stacking faults, and linear defects, such as dislocations, which arealigned so that they are parallel with the incident beam direction. The defects shouldalso be periodic through the entire projected structure. Inclined faults and curved


grain boundaries will not in general be amenable to fruitful study by high-resolutionimaging (nor any other imaging technique). Despite these various constraints, therehave been many, many successful studies where the HREM has proven indispensablein evaluating nanostructured materials. The highly selective group of examples brieflydescribed below is chosen to be illustrative of the myriad possibilities. The need forcomplementary information from macroscopic measurements, including electrical andoptical properties, as well as other types of microscopy, should not be overlooked.

2.6. Milestones

The performance of the “high” resolution electron microscope first exceeded thatof the optical microscope in the mid-1930s [43], but direct correspondence betweenlattice images and projected crystal structure of large-unit-cell block oxides was notachieved for many years [44, 45]. Improved instrumentation eventually led to directlyinterpretable information about atomic arrangements for metals, ceramics and semi-conductors. Individual atomic columns in a small gold particle were resolved [46],and several high-voltage HRTEMs surpassed the 0.2 nm interpretable resolutionlimit on a routine basis in the early 1980s [47–50]. Intermediate-voltage HRTEMsthat could attain this performance level regularly also became available commercially[31, 51, 52]. The highly coherent illumination available with 300-kV FEG TEMsfacilitated the achievement of instrumental resolutions closely approaching 0.1 nm[53], and image interpretation to the same level was achieved by through-focal seriesreconstruction [26] and off-axis electron holography [38]. The latest generation ofhigh-voltage (1–1.5 MV) HRTEMs have succeeded in closely approaching the long-sought-after goal of 0.1 nm without reversals in the contrast transfer function [54–56],enabling direct image interpretation without the need for a posteriori image processing.Significant information transfer beyond the first zero crossover of 0.105 nm was clearlydemonstrated by the 1.25-MeV HRTEM in Stuttgart [57]. Sub-Ångstrom electronmicroscopy to resolutions of better than 0.09 nm has since been achieved using exit-wave retrieval [58, 59] and also using aberration-corrected annular-dark-field imagingwith a scanning transmission electron microscope [60].

3. APPLICATIONS OF HRTEM

Imaging with the HRTEM enables individual atomic columns to be resolved in mostinorganic materials, making it possible to determine the atomic-scale microstructureof lattice defects and other inhomogeneities. Structural features of interest includeplanar faults such as grain boundaries, interfaces and crystallographic shear planes,linear faults such as dislocations and nanowires, as well as point defects, nanosizedparticles and local surface morphology. Additional information can be extracted fromhigh-resolution studies, including unique insights into the controlling influence ofstructural discontinuities on a range of physical and chemical processes such as phasetransformations, oxidation reactions, epitaxial growth and catalysis. The HRTEM hasimpacted many scientific disciplines, and the technique has generated a vast scientificliterature, far too extensive to be reviewed in any detail. Our intent here is to describe,


Figure 3. (a) HRTEM micrograph showing core structure of symmetrical Lomer edge dislocation atGe/Si(001) heterointerface, as recorded with 1.25-MeV atomic-resolution electron microscope;(b) corresponding structural model [65].

albeit very briefly, some representative examples that illustrate the range of applications,while additional information and many more examples can be found elsewhere [1–13].

3.1. Semiconductors

The characterization of semiconductors has been a highly active and fruitful areaof research for HRTEM. Because of the relatively large unit-cells of elemental andcompound semiconductors, ranging from 0.54 nm (Si) to 0.65 nm (CdTe), high-resolution lattice-fringe images are easily obtained in the <110> orientation. However,individual atomic columns are not separately resolved unless the microscope resolutionallows {004}-type reflections to contribute to the imaging process [61]. True atomicimaging for elemental Si and Ge was demonstrated in <100>, <111> and <013>

orientations under carefully chosen imaging conditions using a 400-keV HRTEM[62]. Discrimination between atomic species on the basis of image spots of differentintensity can be achieved for compound semiconductors with sufficiently differentatomic numbers, which is particularly useful for analysing interfaces between dissimilarmaterials.

Dislocations: Individual atomic columns have not actually been resolved in almostall HRTEM structural studies so far reported. Extensive comparisons with simulatedimages for various models of Lomer edge dislocation cores in Si and Ge enabledthe locations of atomic columns to be deduced to within ∼0.025 nm [63]. HRTEManalysis of dissociated 60◦ dislocations in CdTe showed that alternative structural models(glide set and shuffle set) could be clearly differentiated, but again without resolutionof individual atomic columns [64]. By taking advantage of the improved microscoperesolution that has recently become available, more accurate modelling of dislocationcore structures in semiconductors should nowadays become possible. Figure 3 showsan example, recorded with the 1.25-MeV HRTEM in Stuttgart, of a symmetricalLomer edge dislocation at a Ge/Si(001) heterointerface, together with the structuralmodel derived directly from the HRTEM image appearance [65].


Figure 4. Cross-section of CoSi2 nanowire grown by reactive epitaxy on Si(100) surface at 750◦C [74].

Interfaces: Many semiconductors have grain boundaries (GBs) and heteroepitaxialinterfaces that are almost atomically flat and can be aligned edge-on to the incidentbeam direction, with the crystals on both sides of the boundary oriented to a low-indexzone axis. The Ge (130) � = 5 GB was studied in [001] and [013] projections, so thata complete three-dimensional crystallographic analysis could be performed [66]. TheGe (112) � = 3 GB was shown to have a c(2 × 2) periodic supercell of the geometricalcoicidence lattice of the boundary [67], and the rotation-twin nature of a GaP � = 3{111} twin boundary was unambiguously identified from the characteristic imagefeatures [57].

Heterointerfaces are of immense practical importance. Interface roughness can bemade apparent by choosing appropriate thickness and defocus values to emphasizecontrast differences between materials, and composition gradients can be assessed fromthe abruptness with which the two characteristic contrast motifs terminate at theinterface. For materials with large lattice mismatch, significant image features canusually be interpreted without resorting to special defocus/thickness combinations.Examples of large misfit systems where the atomic structure of interfacial misfit dis-locations have been successfully determined include the GaAs/Si(001) interface [68],the CdTe(001)/GaAs(001) interface [69], and the CdTe(111)/GaAs(001) interface [70].For silicide-silicon interfaces, the atomic facetting of asymmetrical twins and asymmet-rical “hetero”-twins was investigated [71], and structural studies have been reportedfor various CoSi2 (A, B)/Si(111) interfaces [72, 73]. Figure 4 shows an interestingexample of a CoSi2 nanowire formed by self-assembled epitaxial growth on Si(100)substrate at 750◦C [74].

The Group III-nitrides of AlN, GaN and InN have potential applications in shortwavelength optoelectronic devices, based on their wide bandgaps which range from1.9 eV (InN), to 3.4 eV (GaN), to 6.2 eV (AlN), and their excellent thermal prop-erties make them ideal candidates for high-temperature and high-power devices [75].However, the lack of substrate materials that are both lattice- and thermally-matched


Figure 5. High-resolution electron micrograph showing cross-section of GaN/SiC interface. Analysis oflattice spacings along planes labeled a, b, and c confirmed interface abruptness [76].

represents a serious obstacle to ongoing research. Thus, there is much interest inunderstanding the atomic structure at the substrate-nitride interface. Figure 5 shows ahigh-resolution image of an GaN/SiC interface, recorded at the optimum defocus sothat atomic-column projections have black contrast. Image analysis established that theGaN/SiC interface was atomically abrupt (between the planes labelled a and b), andthat atomic arrangements across the interface primarily consisted of N bonded withSi, but with some Ga bonded with C in order to maintain charge balance [76].

3.2. Metals

Defects in metals pose greater challenges for characterization using HRTEM becauseof reduced unit-cell dimensions. Atomic imaging is restricted to comparatively fewlow-index zone axes even with the latest generation of instruments. High-resolutionimaging requires very thin foils but substantial atomic rearrangements, especially relax-ation in the vicinity of lattice defects and crystal surfaces, are liable to alter the defectstructure so that it becomes atypical.

Grain boundaries and interfaces: High-resolution observations of symmetrical Au<110> tilt GBs established the presence of different recurring structural units [77].Structural characterization of an Al [100] 45◦ twist plus 17.5◦ tilt GB revealed thatthis asymmetrical GB was composed of a mixture of two basic structural units [78].Observations of a Mo bicrystal revealed the presence of both � = 25 and � = 41structural units along the GB [79]. Bi segregation was reported to modify the structureof grain boundaries in Cu [80], and GBs became facetted and an ordered Bi layer wasobserved at a � = 3 twin boundary [81]. Figure 6 shows an example of an Al 6◦ [001]


Figure 6. Atomic-resolution electron micrograph of Al 6◦ [001] symmetric tilt grain boundary withmisfit accommodation by [110]/2 edge dislocations (arrowed). Each black spot corresponds to projectionof individual Al atomic column [82].

symmetric tilt GB with prominent 1/2<110> edge dislocations that accommodatethe tilt misalignment [82].

Structural modelling based upon atomistic simulations has accompanied manyHRTEM GB studies. The incoherent Al {112} twin GB was observed [83], and com-parisons were made with structures calculated using the Embedded Atom Method(EAM). For the Al � = 9 (221)[110] GB, glide-plane and mirror-plane symmetricstructures were observed to alternate periodically along the boundary: the predictionsof the various atomistic modelling approaches could only be distinguished for one ofthese two structures but not the other [84]. The atomic structure of the Nb (310) twinboundary was generated using interatomic potentials derived from several approachesbut only one, the so-called model-generalized pseudopotential theory, successfully pre-dicted the mirror symmetry later observed in experimental images [85]. For the Nb� = 25 (710)/[001] twin, a multiplicity of stable, low-energy structures were predictedby EAM: four possibilities out of 13 remained after careful comparison with exper-imental micrographs [86]. Molecular dynamics simulations of the Ag � = 3<110>

(211) twin boundary predicted a thin boundary phase having the rhombohedral9R structure, and this prediction was confirmed by experimental HRTEM micro-graphs [87]. Figure 7 is an experimental image showing the existence of the samerhombohedral phase at an incoherent Cu � = 3 (211) boundary [88].


Figure 7. Atomic-resolution electron micrograph of Cu � = 84◦ grain boundary showing presence ofpredicted rhombohedral 9R phase located between large-angle grain boundary at right (structural unitsoutlined) and small-angle boundary at left. Black spots represent positions of individual Cu atomiccolumns [88].

Atomic-level structural investigations of intermetallic alloys and several metal/metalsystems have been reported. For example, the NiAl � = 5 [001](310) GB was imagedby high-resolution electron microscopy and analyzed with the assistance of imagesimulations [89]: changes in local stoichiometry and a rigid body translation along butnot normal to the boundary were inferred. A quantitative study of the NiAl � = 3 (111)GB enabled atomic positions at the boundary core to be determined with an accuracyof ∼0.015 nm [90].

Dislocations: Determination of the atomic structure of dislocation cores is simplifiedby the presence of only one atomic species but the core may not be stable in metal foilsthin enough for atomic-resolution viewing due to the possibility of core spreadingor even defect motion in the form of glide to the thin foil edge [91]. Moreover,structural rearrangements during high-resolution observations of thin metal foils havebeen observed by many workers (see, for example, [92]). The 1/3<1120>{1010}edge dislocation in α-Ti was determined to have a planar elongated core structure[93]. Atomic modelling of the core structure of a<100> and a<110> dislocationsin the intermetallic alloy NiAl revealed that the former had large elastic strain fieldsbut were usually undissociated, whereas the latter either decomposed into other typesof dislocations or climb-dissociated into two partial dislocations [94]. An investigationof dislocation core structures in the ordered intermetallic alloy TiAl has also beenreported [95].

3.3. Oxides and Ceramics

Most ceramic materials consist of two or more atomic species and only rarely can theseparate elements be discriminated in HRTEM micrographs. The image of the 6Hpolytype of SiC in the (1120) orientation shown in Fig. 8 is a particularly noteworthy


Figure 8. Atomic-resolution electron micrograph of 6H SiC polytype recorded with 1250-keV HRTEMinstalled in Tokyo. Image simulation (inset) confirms separation of individual Si and C atomic columns[55].

example since the closest separation of the clearly resolved Si and C columns is only0.108 nm [55]. Structure imaging in the electron microscope originated with large-unit-cell block oxides, although the image interpretation was initially based on con-trast features corresponding to the location of tunnels [45]. Improvements in HRTEMresolution to better than 0.2 nm led to oxide images in which the black spots vis-ible at optimum defocus were interpretable in terms of atomic column positions.It is then possible to deduce directly the detailed atomic structure of complicatedshear defects and precipitation phenomena observed in doped and slightly reducedoxides, as represented by the example of a pentagonal bipyramidal defect as shown inFig. 9 [96].

Grain boundaries. Symmetric tilt GBs in NiO bicrystals had a multiplicity of dis-tinct structural units [97], and asymmetric structural models based on experimentalmicrographs, differed from symmetric models derived from theoretical considerations.Observations of the � = 5 (210)/[001] symmetric tilt GB in yttrium aluminum garnetwere compared with model atomic structures [98]. Structural investigations of a near-�= 5 (210) GB in TiO2, rutile, revealed a stepped boundary with well-defined lattice dis-locations at the steps, and extended, flat terraces that consisted of � = 5 (210) segmentswith mirror glide symmetry [99]. Observation of an undoped SrTiO3 � = 5 (130)symmetrical tilt GB revealed that it was composed of repeating structural units [100].Examination of a 25◦ [001] tilt boundary in SrTiO3 provided an initial structural model,and bond-valence sum calculations based on electron-energy-loss spectroscopy at theboundary were then used to refine the O atom positions [101]. A combined structural


(a) (b)

Figure 9. (a) Atomic-resolution electron micrograph of nonstoichiometric (W, Nb)O2.93 showing pairsof pentagonal bipyramidal columnar defects. (b) Corresponding structural model. Occupied tunnel sitesare located by direct visual inspection [96].

and spectroscopic investigation of coherent (111) twins in BaTiO3 enabled a modifiedstructural model of the boundary to be proposed [102]. The atomistic structure of90◦ domain walls in ferroelectric PbTiO3 thin films was investigated by using digitalprocessing to determine lattice parameter variations across the walls [103].

Interfaces. Knowledge of metal-ceramic interfaces promotes a better understandingof bulk mechanical properties. For thin layers of Nb and Mo deposited on R-planesapphire, misfit dislocations were offset from the interface for Nb films whereas theywere localized very close to the interface for Mo films [104]. Interactive digital imagematching was used in a comprehensive study of the Nb/sapphire interface: and trans-lation vectors were determined with a precision of ∼0.01 nm [105]. The atomic corestructure of the dislocations was determined in another study of the same interface[106].

Ceramics. Most ceramics are close-packed materials with tetrahedral bonding so thatatomic resolution is very difficult to obtain. Careful examination of {100} nitrogenplatelets in diamond led to the development of a novel “nitrogen-fretwork” model[107], while later observations led to a more refined model [108]. Interfacial struc-tures and the defects that occurred during heteroepitaxial growth of β-SiC films onTiC substrates have been characterized [109]. The atomic structure of Ti(C,N)-TiB2

interfaces was investigated using a combined microscopy-simulation study [110].

3.4. Surfaces

Several different TEM configurations provide atomic-scale information about surfaces[111]. In surface profile imaging [112], the electron microscope is operated in thenormal HRTEM mode and the optimum defocus image displays the surface profile atthe structural resolution limit. Gold received much early attention, mainly because itssurface was relatively inert and electron irradiation readily caused desorption of surface


Figure 10. Surface profile images showing reconstructed CdTe (001) surface at different temperatures:(a) 2 × 1 at 140◦C; (b) 3 × 1 at 240◦C [127].

contamination overlayers, and also because its large interatomic spacings and highatomic-column visibility made it easy to characterize atomic rearrangements [113].Thus, profile imaging was first applied to observations of a 2 × 1 reconstruction of thegold (110) surface [114]. Adsorbed Bi atoms have been identified on reconstructedSi(111) surfaces using an ultrahigh-vacuum (UHV) transmission electron microscope[115], and the overgrowth of Au on ZnTe has been investigated [116]. In later exper-iments involving in situ Au evaporation under UHV conditions, the ×5 superperiodassociated with the corrugated (5 × 28) reconstructed Au surface was observed [117].

Oxide surfaces are generally more straightforward to prepare for profile imaging[118], but surface modification may occur under intense electron irradiation due toelectron-stimulated desorption of oxygen from near-surface regions [119, 120]. Acomplex spinel catalyst developed surface rafts, identified as ZnO, following prolongeduse as an oxidation catalyst [121]. Surface profile images were central to studies ofterbium oxide [122], Eu2O3 [123], and β−PbO2 [124]. A direct correlation wasmade between the exposed surface structure of V2O5 oxide catalysts and the catalystselectivity [125].

Profile images of semiconductor surfaces can only really be considered as valid whenthe surface has been cleaned inside the microscope. A novel 1 × 1 dimer reconstructionof a Si (111) surface was reported after the sample had been heated in situ to 1000◦C[126]. As shown by the surface profile images in Fig. 10, the CdTe (001) surfaceundergoes a reversible phase transformation from a 2 × 1 structure at temperaturesbelow about 200◦C to a 3 × 1 structure at higher temperatures [127]. The atomiccolumns at the surface were located to within ∼0.01 nm, and structural models were


developed: the 2 × 1 was determined as being Cd-stabilized whereas the 3 × 1 wasfound to be Te-stabilized.

3.5. Dynamic Events

Dynamic events such as phase transitions, defect motion and interface dynamics canbe documented directly at the atomic-scale by means of a low-light-level TV cameraattached to the base of the microscope without any loss of resolution. However, imagerecording is not useful for quantitative image analysis due to reduced dynamic responseof the camera, as well as image distortions. Further information about dynamic in situstudies can be found in the chapter by Sharma and Crozier.

Surfaces. Surface profile imaging revealed movement of Au atomic columns acrossextended Au (110) surfaces [113], and rapid structural changes and the existence of“atom clouds” extending out from Au surfaces were reported [128]. Hopping ofAu columns between surface sites on small Au particles was recorded at TV rates[129, 130], and similar hopping of Pt columns was also studied [131].

Small particles. Under strong electron irradiation, small metal particles (<8–10 nm)rapidly change their shape and orientation. Structural rearrangements from single crys-tal twinned to multiply-twinned were observed in small Au particles [130, 132]. Struc-tural rearrangements as well as surface hopping were also documented in small particlesof Pt [131] and Rh [133]. In the case of small Ru particles (∼2.5 nm), the internalstacking changed between cubic-close-packed (ccp) and hexagonal-close-packed (hcp)which is the stable bulk form of Ru [134]. The term “quasi-melting” has been used todescribe the structural fluctuations which have been reported for Au clusters supportedon pillars of MgO [135].

4. CURRENT TRENDS

4.1. Image Viewing and Recording

Image viewing or recording should not normally be expected to affect resolution limitsof the HRTEM but the recording media must still be properly optimized to ensureefficient operation. Ideally, every incident electron should be detected but readout noiseand shot noise could affect the overall recording efficiency. The low-light-level TVcamera has steadily evolved to the point where it has replaced the fluorescent screenfor most image viewing. A camera is easily attached beneath the viewing chamberwithout affecting microscope performance, and high brightness images can be obtainedeven for very low exposure levels. Enlarged specimen detail is easily visible on a TVmonitor, and permits image focussing and rapid astigmatism correction, while dynamicevents within the sample can be viewed and recorded at TV rates. Nevertheless, highdark current and amplifier noise limit dynamic range, the input-output linearity ispoor, and the number of resolvable picture elements is restrictive. The intensifiedTV camera is seriously inadequate as a recording medium for quantitative HRTEMstudies.

The development of the slow-scan CCD camera has led to a revolution acrossthe entire field of electron microscopy, and it has particular value for quantitative


HRTEM applications. The potential of the CCD camera had long been recognized[136], especially its sensitivity, wide dynamic range and overall usefulness for theelectron microscopist [137]. However, dynamic viewing at TV rates seems unlikely tobe achieved without sacrificing image quality. The imaging properties of the slow-scanCCD camera, the intensified TV camera and the photographic plate have been com-pared [138], and a comprehensive overview of characterization methods and importantdetection parameters, such as modulation transfer function and input-output linear-ity, for CCD cameras has been published [139]. The fixed location of the CCDcamera enables geometric distortions of the imaging system to be accurately com-pensated [140], which is advantageous for extracting quantitative phase informationduring off-axis electron holography. On-line acquisition of a digital signal also enablesautomated microscope control or “autotuning” [137], as described in the followingsection.

4.2. On-Line Microscope Control

As resolution limits improve, it becomes progressively more difficult for an operator toadjust focus, to correct the objective lens astigmatism and to align the incident beamdirection (coma-free alignment) with the accuracy that is required to ensure that imageinterpretation is not compromised. Accordingly, attention has been directed towardson-line computer control or “autotuning” of the microscope [137, 141]. The desiredend-product of routine, top-quality micrographs should then leave the microscopistfree to concentrate on solving the particular materials problem at hand.

Several criteria have been proposed as a basis for autotuning, including diffrac-tograms, beam-tilt-induced image displacement, and contrast analysis for amorphousmaterials [24]. These methods rely upon signals that are fed into a computer, whichanalyses the data and then makes appropriate on-line adjustments to the microscopecontrols. Contrast analysis, the first autotuning method to be implemented successfully[141], locates a global minimum of the image variance as the computer successivelyiterates through the focus, beam tilt controls (in two orthogonal directions) and twoobjective lens stigmator controls. High accuracy for beam alignment and astigmatismcan be achieved but the method is inefficient in terms of dose and inapplicable in theabsence of amorphous material.

The latest variant of autotuning to be implemented [137], which is based on auto-mated diffractogram analysis (ADA), utilizes diffractograms computed from slow-scanCCD images of amorphous material. The method works best at high image magni-fication, but is again inapplicable in the absence of an amorphous specimen region.The set of diffractograms in Figure 11, recorded (a) before, (b) after one cycle, and(c) after two cycles, of autotuning, illustrate the improvements that can be obtainedwith a 200-keV FEG-TEM. Astigmatism correction and focus adjustment to within1 nm can be achieved, and the beam-tilt alignment is better than 0.1 mrad, which isbeyond the adjustment/detection limit of the available instrumentation. Autotuningshould be regarded as an essential procedure during preparation of the microscope forHRTEM observation.


Figure 11. Diffractogram tableaus from thin amorphous carbon film recorded (a) before; (b) after onecycle; (c) after two cycles, autotuning sequence based on automated diffractogram analysis (ADA). Initialtilt misalignment reduced successively from 4 mrad (a) to 0.4 mrad (b) and <0.1 mrad (c).

4.3. Detection and Correction of Third-Order Aberrations

All electron lenses suffer from performance-limiting aberrations that must be correctedwhen possible, or at least quantified and accounted for during image interpretation.Two-fold astigmatism and third-order axial coma can be corrected during autotuningas discussed in the previous section. Detection and correction of spherical and chro-matic aberration, as well as three-fold objective-lens astigmatism become much morecritical as microscope resolution limits extend towards and beyond the 0.1-nm barrier[142–144].

Spherical aberration is unavoidable in rotationally symmetric electron lenses[145, 146]. Elimination of Cs (and Cc ) by a suitable combination of multipole ele-ments attracted much attention over many years but all early correction attempts failed,due primarily to insufficient electrical stability and lack of alignment precision [146].In recent years, a double-hexapole Cs -corrector system attached to a 200-keV FEG-TEM has enabled the normal 0.23 nm interpretable resolution limit of the instrumentto be surpassed, with a level of ∼0.13 nm eventually being achieved after adventi-tious instabilities were removed [147]. Concurrently, a corrector system, incorporat-ing multiple quadropole-octopole elements, has been applied to the probe-forminglens of a 100-keV scanning transmission electron microscope, and probe sizes of lessthan 0.1 nm can be achieved [148]. Note that both approaches to aberration cor-rection are completely dependent on computer analysis of the imaging conditionsand high-precision feedback to the numerous deflector and corrector power supplies.Some initial experiences with aberration-corrected HRTEM are described later inSection #4.5.

Knowledge of the chromatic aberration coefficient of the objective lens is not criticalfor high-resolution imaging since Cc does not affect the interpretable resolution, andthe temporal coherence envelope is determined by an effective focal spread, which canbe estimated empirically if required [23]. Nevertheless, Cc does impact temporal coher-ence, which is performance-limiting for an LaB6 electron source, so that reduction


or correction of chromatic aberration is desirable and worth pursuing. Correction ofCc in a low-voltage scanning electron microscope was achieved [149] but no substan-tial success in the energy range applicable for HREM (upward from about 100 kV)has so far been reported. An alternative approach to reducing chromatic effects is tolocate a monochromator immediately following the electron gun, which could serveto reduce the energy spread to about 0.1 eV [150], but this improvement can only beachieved at the expense of reduced beam current [151].

Three-fold astigmatism was quantified during the first comprehensive study ofcoma-free alignment [40]. However, it was not until the implementation of the ADAmethod for autotuning that the disturbing implications of three-fold astigmatism forhigh-resolution imaging at the 0.10 nm level came to be fully appreciated [41]. Likeaxial coma, three-fold astigmatism is effectively invisible in an axial bright-field imageor the corresponding diffractogram since its basic effect is an asymmetrical shift ofphase information. However, image simulations have shown that the impact on very-high-resolution images of crystalline materials can be highly detrimental, dependingon the relative orientation of the three-fold astigmatism and any crystal symmetrydirections [143, 144]. The magnitude of three-fold astigmatism can be estimated usingdiffractograms from four mutually orthogonal beam-tilt directions, and correction canbe achieved using a pair of sextupole stigmator coils located near the back focal planeof the objective lens: Reasonable adjustments can be reached using fixed correctioncurrents in existing objective stigmator coils provided that the coils are energizedseparately rather than being wired in pairs [152]. Microscopists purchasing new micro-scopes should insist that this highly desirable correction of three-fold astigmatism bedone during installation and commissioning of their instrument.

4.4. Quantitative HRTEM

A major attraction of HRTEM is the possibility that atomic arrangements at localirregularities such as dislocations and interfaces can be determined to very high accu-racy, some times closely approaching 0.01 nm. However, the refinement process isheavily demanding both of the microscopist and the instrument, as well as beingcomputing-intensive and often extremely time-consuming with much trial-and-errorparameter fitting. Prior knowledge (or elimination) of essential experimental parame-ters facilitates the goal of interactive structure refinement. Correction of the three-foldastigmatism at the time of microscope installation should reduce its effect to the levelwhere it has almost negligible influence on the image. Routine application of auto-tuning, which implies the availability of an online CCD camera as well as computercontrol of relevant power supplies, would remove two-fold astigmatism (two parame-ters) and coma/beam tilt (two more parameters) from consideration. Thus, additionalresources must be committed in order to make progress towards the ultimate goal ofreal-time, interactive structure refinement. For aperiodic structural features, it appearsthat this goal will not be easily attained since some parameters can only be optimizediteratively [105], although it is possible to automate considerably parts of the iterationprocess [153].


Historically, the determination of defect structures has relied upon qualitative com-parisons between experimental micrographs and image simulations that were based onvarious alternative structural models. The acceptability of a specific structural modelwas generally considered as being enhanced when an image match was achieved formultiple members of a focal series (see, for example, refs. 85, 107). An alternativeapproach has been to overlay projected atomic column positions on the experimentaland/or simulated images [66], and other studies have utilized superposition [79] orsubtraction [154] of simulated and experimental images during refinement. A non-linear least-squares optimization approach was used to refine atomic positions at NbGBs [155].

Attention has been given to the issue of quantifying the “goodness of fit” betweenexperimental micrographs and the corresponding simulated images, as derived frompostulated structural model(s). The so-called reliability or R-factors of X-ray or neu-tron diffraction studies refer to the contents of the entire unit cell, unlike aperiodicdefects such as grain boundaries or dislocation cores studied by HRTEM where suchwell-defined discrete entities do not generally exist. Thus, alternative image agreementfactors (IAFs) have been proposed and utilized at different times, and a useful summaryof those most commonly used can be found in the Appendix of ref. [105]. However, itshould be noted that slightly different results can be obtained depending on which IAFis used, possibly because different IAFs weight bright and dark contrast areas differently.It would be helpful if defocus and other microscope parameters were removed from theactual refinement process by instead using the complex exit-surface wavefunction ofthe specimen as the basis for comparison between simulation and experiment. At thatstage, a χ2 goodness-of-fit criterion would be an appropriate test of overall conver-gence [156]. An additional benefit of determining the exit-surface wavefunction wouldbe the availability of both phase and amplitude information about electron scatteringby the sample. The possibilities for determining unknown object structures would beenhanced because this wavefunction directly reflects electron scattering by the object.

4.5. Aberration-Corrected HRTEM

Compensation of the spherical aberration of the objective lens offers the excitingprospect of directly interpretable image detail extending out to the HRTEM infor-mation limit without the need to unscramble the artefactual detail normally caused byTF oscillations. One additional benefit of the aberration-corrected HRTEM is thatimage delocalization, which is a major source of imaging artefacts at discontinuitiessuch as interfaces and surfaces, is markedly reduced [147, 157]. Another benefit isthat other imaging aberrations are also substantially reduced, which should simplifythe process of exit-wave retrieval using through-focal reconstruction [158], and alsoalleviate the accuracy needed for sample tilting. However, it is relevant to reiteratehere that conventional TF theory for the HRTEM refers to phase contrast imaging,so that when the Cs value is reduced exactly to zero it would be necessary to use aprojected charge density approach for image interpretation [159]. In practice, havingan adjustable spherical aberration can provide additional flexibility to the microscopist


interested in solving particular materials problems. For example, a small negative Cs

with a slightly overfocus condition enabled imaging of oxygen atom columns in aperovskite ceramic [160]. It is clear that much experimentation is still required toexamine the full range of possibilities for aberration-corrected HRTEM, and possiblyreach some consensus about standard imaging conditions.

5. ONGOING PROBLEMS

5.1. The Stobbs’ Factor

It has become increasingly obvious, and highly disconcerting, that there are substantial,seemingly inexplicable, discrepancies between the contrast levels of experimental andsimulated images as well as diffracted beam intensities [161]. These differences werenot apparent in earlier qualitative studies using photographic film when there was nosimple measure of absolute intensity, and the contrast range in image simulations couldeasily be scaled to match that of the experimental micrographs. Initial quantitativestudies revealed major differences in image contrast, sometimes by factors as large as6 to 8 [161], although factors of about three are reported to be more typical [162]Obvious sources of error, such as contributions from inelastic scattering and surfacecontamination overlayers, have not adequately accounted for these contrast differences[163], stimulating further concerted efforts to identify the origin(s) of what has cometo be called the Stobbs’ factor [164]. Thermal diffuse scattering has been at least partlyimplicated by recent experiments combining energy-filtered imaging with off-axiselectron holography [165], leading to suggestions for further experiments that mightfinally be definitive [164].

5.2. Radiation Damage

Interactions between the highly energetic electron beam and the sample within theelectron microscope are always likely to result in permanent structural modification.There are two basic types of electron beam damage [166]: radiolytic processes (some-times known as “ionization damage”) involve electron-electron interactions and affectmost covalent and ionic solids; and direct atomic “knock-on” displacements, whichoccur above characteristic energy thresholds. An electron energy of 400 keV is suffi-cient to cause bulk displacements in elements as heavy as copper (Z = 29). Moreover,because of reduced binding energy, the energy needed for the activation of surfacesputtering or for diffusion at defects or interfaces is considerably less than bulk values.For example, under continuous electron irradiation (400-keV, 5–15 A/cm2), electron-stimulated desorption causes depletion of oxygen from the near-surface region ofmaximally valent, transition-metal oxides, leaving thin layers of reduced oxide cover-ing the surface [119]. In a quantitative comparison of difference images, preferentialdamage was documented to occur at a Cu/sapphire interface [167], which limited theuseful viewing time to a total of ∼10 mins (1250 keV, magnification of 600,000×,specimen current density of ∼1.6 A/cm2).

Thus, the electron microscopist must be continually alert to the likelihood thatthe sample morphology has altered, probably irreversibly, during HRTEM imaging


and especially microanalysis. Higher viewing magnifications as a means to improvesignal statistics result in higher current densities, which will mean higher damage ratessince the current density at the sample increases with the square of the magnificationwhen the screen brightness is maintained at a constant level. For quantitative studies,the image magnification and the beam current density should be limited wheneverpossible, and the region of sample studied must be periodically checked for signs ofstructural change. A cautious microscopist will monitor the sample appearance duringobservation and thus erroneous results can be discounted once changes start to becomeapparent [168]. Digital micrographs of an NiAl � = 3 (111) twin boundary, recordedwith a slow-scan CCD camera in a high-voltage HREM, were compared at regularintervals as a means of establishing the useful observation time [90]. Finally, it should beapparent that structural change is even more likely to occur when the highly intense,focussed probe of the STEM is used for nanoscale microanalysis, and this possibilityshould always be monitored.

5.3. Inversion of Crystal Scattering

Inversion of crystal scattering to retrieve the crystal potential in the presence of dynam-ical scattering is a major unresolved problem. Several methods exist for retrieving theexit-surface wavefunction, and Fourier inversion can be used to extract the crystalpotential for very thin samples when the kinematical or weak phase object approxima-tions are valid. However, ab initio inversion of crystal scattering to retrieve the crystalpotential has received comparatively scant attention over the years. An iterative methodbased on inversion of the multislice algorithm has been proposed [169], but furtherwork is still needed to explore the applicability of the method, and for extendingthickness limits in the case of non-periodic wavefields [170]. An alternative approachbased on a simulated annealing algorithm has been explored for a perfect GaAs crystalwith zincblende structure [171]. Successful reconstruction was achieved for a thicknessof 5.6 nm but not for a thickness of 11.2 nm. Overall, it still holds true that becauseof the likelihood of multiple solutions to the inverse scattering problem for almost anysample of reasonable thickness, the uniqueness of the inversion process for unknownstructures remains an unresolved issue.

6. SUMMARY AND FUTURE PERSPECTIVE

This chapter has provided an overview of HRTEM, with the objective of highlight-ing some of its applications and achievements, as well as identifying areas of ongoingresearch and development. The HRTEM enables the atomic structure of interfacesand defects to be determined routinely, reliably and with very high positional accu-racy, thus providing better insights into the physical behavior of many nanostructuredmaterials. Experts in the HRTEM field should find ways to ensure that their colleagueswho are interested in the properties of these types of materials are given the oppor-tunity and assistance required to capitalize on its attractive possibilities. Meanwhile,further developments in quantification and many novel applications can be antic-ipated. Online microscope control (“autotuning”), digital recording and computer


processing, (almost) real-time structure refinement, and in situ environmental electronmicroscopy are likely areas of concentrated activity. Many challenges remain. The dif-ferences between simulated and experimental contrast levels need to be fully explained.Better approaches to inversion of crystal scattering are needed. Operating conditionsfor aberration-corrected imaging need to be further explored. These issues will surelyreceive much attention over the next several years.

REFERENCES

1. D. J. Smith, Rep. Prog. Phys. 60 (1997) 1513.2. F. Ernst and M. Ruhle, Current Opin. Solid State & Mats. Sci. 2 (1997) 469.3. J. C. H. Spence, Mater. Sci. Eng. R26 (1999) 1.4. Z. L. Wang, Adv. Maters. 15 (2003) 1497.5. W. Krakow, F. A. Ponce, and D. J. Smith (Eds.), High Resolution Microscopy of Materials, MRS Symp.

Proc. Vol. 139, Materials Research Society, Pittsburgh (1989).6. R. Sinclair, D. J. Smith, and U. Dahmen (Eds.), High Resolution Electron Microscopy of Defects in Materials,

MRS Symp. Proc. Vol. 183, Materials Research Society, Pittsburgh (1990).7. D. Biegelsen, D. J. Smith, and S.-Y. Tong (Eds.), Atomic Scale Imaging of Surfaces and Interfaces, MRS

Symp. Proc. Vol. 295, Materials Research Society, Pittsburgh (1993).8. D. J. Smith and J. C. H. Spence (Eds.), Aspects of Electron Microscopy, Diffraction, Crystallography and

Spectroscopy Ultramicroscopy 52, Nos. 3/4, December 1993.9. M. Ruhle, F. Phillipp, A. Seeger, and J. Heydenreich (Eds.), High-Voltage and High-Resolution Electron

Microscopy Ultramicroscopy 56, Nos. 1–3, November 1994.10. M. Sarikaya, H. K. Wickramasinghe, and M. Isaacson (Eds.), Determining Nanoscale Physical Properties

of Materials by Microscopy and Spectroscopy, MRS Symp. Proc. Vol. 332, Materials Research Society,Pittsburgh (1995).

11. J. C. H. Spence, Experimental High Resolution Electron Microscopy, Clarendon, Oxford (1988).12. S. Horiuchi, Fundamentals of High Resolution Transmission Electron Microscopy, North Holland, Amsterdam

(1994).13. D. Shindo and K. Hiraga, High Resolution Electron Microscopy for Materials Science, Springer, Tokyo (1998).14. P. G. Self and M. A. O’Keefe, In: High-Resolution Transmission Electron Microscopy and Associated Techniques,

P. R. Buseck, J. M. Cowley, and L. Eyring (Eds.), Chapter 8, Oxford University Press, New York (1988).15. D. Van Dyck, J. Microscopy 132 (1983) 31.16. J. M. Cowley and A. F. Moodie, Acta Cryst. 10 (1957) 609.17. P. Goodman and A. F. Moodie, Acta Cryst. 30 (1974) 280.18. P. A. Stadelmann, Ultramicroscopy 21 (1987) 131.19. P. G. Self, M. A. O’Keefe, P. R. Buseck, and A. E. C. Spargo, Ultramicroscopy 11 (1983) 35.20. J. M. Cowley, Diffraction Physics, Third Revised Edition, North Holland, Amsterdam (1995).21. K.-H. Hanszen, Adv. Opt. Electron Microsc. 4 (1971) 1.22. J. Frank, Optik 38 (1973) 519.23. W. O. Saxton, Optik 49 (1977) 51.24. D. J. Smith, Adv. Opt. Electron Microsc. 11 (1988) 1.25. M. A. O’Keefe, Ultramicroscopy 47 (1992) 282.26. W. Coene, G. Janssen, M. Op de Beeck, and D. van Dyck, Phys. Rev. Lett. 69 (1992) 3743.27. J. M. Cowley and A. F. Moodie, Proc. Phys. Soc. 70 (1957) 486.28. S. Iijima and M. A. O’Keefe, J. Microscopy 17 (1979) 347.29. D. J. Smith, L. A. Bursill, and G. J. Wood, J. Solid State Chem. 50 (1983) 51.30. K. Izui, S. Furuno, T. Nishida, H. Otsu, and S. Kuwabara, J. Electron Microscopy 27 (1978) 171.31. J. L. Hutchison, T. Honda, and E. D. Boyes, JEOL News 24E (1986) 9.32. D. J. Smith, R. A. Camps, and L. A. Freeman, Inst. Phys. Conf. Ser. 61 (1982) 381.33. M. A. O; Keefe, U. Dahmen, and C. J. D. Hetherington, Mater. Res. Soc. Symp. Proc. 159 (1990) 453.34. A. R. Wilson, A. E. C. Spargo, and D. J. Smith, Optik 61 (1982) 63.35. A. Thust and K. Urban, Ultramicroscopy 45 (1992) 23.36. W. E. King and G. H. Campbell, Ultramicroscopy 56 (1994) 46.37. O. L. Krivanek, Optik 45 (1976) 97.38. A. Orchowski, W. D. Rau, and H. Lichte, Phys. Rev. Lett. 74 (1995) 399.


39. W. M. J. Coene and T. J. J. Denteneer, Ultramicroscopy 38 (1991) 235.40. F. Zemlin, K. Weiss, P. Schiske, W. Kunath, and K.-H. Herrmann, Ultramicroscopy 3 (1978) 49.41. O. L. Krivanek, Ultramicroscopy 55 (1994) 419.42. D. J. Smith, W. O. Saxton, M. A. O’Keefe, G. J. Wood, and W. M. Stobbs, Ultramicroscopy 11 (1983)

263.43. E. Ruska, Zeit. f. Physik 85 (1934) 580.44. J. G. Allpress and J. V. Sanders, J. Appl. Cryst. 6 (1969) 165.45. S. Iijima, J. Appl. Phys. 42 (1971) 5891.46. V. E. Cosslett, R. A. Camps, L. A. Freeman, W. O. Saxton, D. J. Smith, W. C. Nixon, H. Ahmed,

C. J. D. Catto, J. R. A. Cleaver, P. Ross, K. C. A. Smith, and A. E. Timbs, Nature 281 (1970) 49.47. M. Hirabayashi, K. Hiraga, and D. Shindo, Ultramicroscopy 9 (1982) 197.48. D. J. Smith, R. A. Camps, V. E. Cosslett, L. A. Freeman, W. O. Saxton, W. C. Nixon, C. J. D. Catto,

J. R. A. Cleaver, K. C. A. Smith, and A. E. Timbs, Ultramicroscopy 9 (1982) 203.49. K. Yagi, K. Takayanagi, K. Kobayashi, and S. Nagakura, In: Proc. Seventh Int. Conf. High Voltage Electron

Microscopy, R. M. Fisher, R. Gronsky, and K. H. Westmacott (Eds.), Lawrence Berkeley Laboratory,Berkeley (1983) p. 11.

50. R. Gronsky and G. Thomas, In: Proc. 41st. Ann. Meet. Electron Microscopy Society of America,G. W. Bailey (Ed.), San Francisco Press, San Francisco (1983) p. 310.

51. D. J. Smith, J. C. Barry, A. K. Petford, and J. C. Wheatley, JEOL News 22E (1985) 2.52. A. Bourret and J. M. Penisson, JEOL News 25E (1987) 2.53. D. van Dyck, H. Lichte, and K. D. vander Mast, Ultramicroscopy 64 (1997) 1.54. S. Horiuchi, Y. Matsui, Y. Kitami, M. Yokoyama, S. Suehara, X. J. Wu, I. Matsui, and T. Katsuta,

Ultramicroscopy 39 (1991) 231.55. H. Ichinose, H. Sawada, E. Takano, and M. Osaki, J. Electron Microscopy 48 (1999) 887.56. F. Phillipp, R. Hoschen, Osaki, G. Mobus, and M. Ruhle, Ultramicroscopy 56 (1994) 1.57. F. Phillipp, Adv. Solid State Physics 35 (1996) 257.58. M. A. O’Keefe, C. J. D. Hetherington, Y. C. Wang, E. C. Nelson, J. H. Turner, C. Kisielowski, J.-O.

Malm, R. Mueller, J. Ringnalda, M. Pan, and A. Thust, Ultramicroscopy 89 (2001) 215.59. C. Kisielowski, C. J. D. Hetherington, Y. C. Wang, R. Kilaas, M. A. O’Keefe, and A. Thust, Ultrami-

croscopy 89 (2001) 243.60. P. E. Batson, N. Dellby, and O. L. Krivanek, Nature 418 (2002) 617.61. R. W. Glaisher, A. E. C. Spargo, and D. J. Smith, Ultramicroscopy 27 (1989) 131.62. A. Bourret, J. L. Rouviere, J. Spendeler, phys. stat. sol. (a) 107 (1988) 481.63. A. Bourret, J. Desseaux, and A. Renault, Phil. Mag. A45 (1982) 1.64. P. Lu and D. J. Smith, Phil Mag. B62 (1990) 435.65. J. N. Stirman, P. A. Crozier, D. J. Smith, F. Phillipp, G. Brill, and S. Sivananthan, Appl. Phys. Lett. (2004)

in press.66. A. Bourret, J. L. Rouviere, and J. M. Penisson, Acta Cryst. A44 (1988) 838.67. A. Bourret and J. J. Bacmann, Surf. Sci. 162 (1985) 495.68. D. Gerthsen, Phil. Mag. A67 (1993) 1365.69. J. E. Angelo and M. J. Mills, Phil. Mag. A72 (1995) 635.70. A. Bourret, P., G. Feuillet, and S. Tatarenko, Phys. Rev. Lett. 70 (1993) 311.71. W.-J. Chen and F.-R. Chen, Ultramicroscopy 51 (1993) 316.72. C. W. T. Bulle-Lieuwma, A. F. de Jong, A. H. van Ommen, J. F. van der Veen, and J. Vrijmoeth, Appl.

Phys. Lett. 55 (1989) 648.73. A. Catana, P. E. Schmid, P. Lu, and D. J. Smith, Phil. Mag. A66 (1992) 933.74. Z. He, D. J. Smith, and P. A. Bennett, Phys. Rev. Lett. (2004) in press.75. H. Morkoc, S. Strite, G. B. Gao, M. E. Lin, B. Sverdlov, and M. Burns, J. Appl. Phys. 76 (1994) 1363.76. J. N. Stirman, F. A. Ponce, A. Pavlovska, I. S. T. Tsong, and D. J. Smith, Appl. Phys. Lett. 76 (2000) 822.77. H. Ichinose, Y. Ishida, N. Baba, and K. Kanaya, Phil. Mag. A52 (1985) 51.78. M. Shamsuzzoha, D. J. Smith, and P. Deymier, Phil. Mag. A64 (1991) 719.79. J. M. Penisson, T. Nowicki, and M. Biscondi, Phil. Mag. A58 (1988) 947.80. D. E. Luzzi, Ultramicroscopy 37 (1991) 180.81. D. E. Luzzi, M. Yan, M. Sob, and V. Vitek, Phys. Rev. Lett. 67 (1991) 1894.82. M. Shamsuzzoha, D. J. Smith, and P. Deymier, Scripta Met. et Mater. 24 (1990) 1611.83. J. M. Penisson, M. J. Mills, and U. Dahmen, Phil. Mag. Lett. 64 (1991) 327.84. M. J. Mills, M. S. Daw, G. J. Thomas, and F. Cosandey, Ultramicroscopy 40 (1992) 247.85. G. H. Campbell, W. L. Wien, W. E. King, S. M. Foiles, and M. Ruhle, Ultramicroscopy 51 (1993) 247.


86. G. H. Campbell, S. M. Foiles, P. Gumbsch, M. Ruhle, and W. E. King, Phys. Rev. Lett. 70 (1993) 449.87. F. Ernst, M. W. Finnis, D. Hofmann, T. Muschik, U. Schonberger, U. Wolf, and M. Methfessel, Phys.

Rev. Lett. 69 (1992) 620.88. D. Hofmann and F. Ernst, Interface Sci. 2 (1994) 201.89. R. W. Fonda and D. E. Luzzi, Phil. Mag. A68 (1993) 1151.90. K. Nadarzinski and F. Ernst, Phil. Mag. A74 (1996) 641.91. M. J. Mills, M. S. Daw, and S. M. Foiles, Ultramicroscopy 56 (1994) 79.92. D. L. Medlin, C. B. Carter, J. E. Angelo, and M. J. Mills, Phil. Mag. A75 (1997) 733.93. A. de Crecy, A. Bourret, S. Naka, and A. Lasalmonie, Phil. Mag. A47 (1983) 245.94. M. J. Mills and D. B. Miracle, Acta metall. mater. 41 (1993) 85.95. K. J. Hemker, B. Viguier, and M. J. Mills, Mats. Sci. Eng. A164 (1993) 391.96. L. A. Bursill and D. J. Smith, Nature 309 (1984) 319.97. K. L. Merkle and D. J. Smith, Phys. Rev. Lett. 59 (1987) 2887.98. G. H. Campbell, J. Amer. Ceram. Soc. 79 (1996) 2883.99. U. Dahmen, S. Paciornik, I. G. Solorzano, and J. Vandersande, Interface Sci. 2 (1994) 125.

100. V. Ravikumar and V. P. Dravid, Ultramicroscopy 52 (1993) 557.101. M. M. McGibbon, N. D. Browning, M. F. Chisholm, A. J. McGibbon, S. J. Pennycook, V. Ravikumar,

and V. P. Dravid, Science 266 (1994) 102.102. A. Recnik, J. Bruley, W. Mader, D. Kolar, and M. Ruhle, Phil. Mag. B70 (1994) 1021.103. S. Stemmer, S. K. Streiffer, F. Ernst, and M. Ruhle, Phil. Mag. A71 (1995) 713.104. D. M. Tricker and W. M. Stobbs, Phil. Mag. A71 (1995) 1037.105. G. Mobus and M. Ruhle, Ultramicroscopy 56 (1994) 54.106. G. Gutenkunst, J. Mayer, and M. Ruhle, Scripta metall. mater. 31 (1994) 1097.107. J. C. Barry, Phil. Mag. A64 (1991) 111.108. P. J. Fallon, L. M. Brown, J. C. Barry, and J. Bruley, Phil. Mag. A72 (1995) 21.109. F.-R. Chien, S. R. Nutt, J. M. Carulli, N. Buchan,C. P. Beetz, and W. S. Yoo, J. Mater. Res. 9 (1994)

2086.110. J. Y. Dai, Y. G. Wang, D. X. Li, and H. Q. Ye, Phil. Mag. A70 (1994) 905.111. D. J. Smith, In: Chemistry and Physics of Solid Surfaces VI, R. Vanselow and R. Howe (Eds.), Springer,

Berlin (1986) Chapter 15.112. D. J. Smith, R. W. Glaisher, P. Lu, and M. R. McCartney, Ultramicroscopy 29 (1989) 123.113. D. J. Smith and L. D. Marks, Ultramicroscopy 16 (1985) 101.114. L. D. Marks and D. J. Smith, Nature 303 (1983) 316.115. Y. Haga and K. Takayanagi, Ultramicroscopy 45 (1992) 95.116. M. Shiojiri, N. Nakamura, C. Kaito, and T. Miyano, Surf. Sci. 126 (1983) 719.117. T. Hasegawa, K. Kobayashi, N. Ikarashi, K. Takayanagi, and K. Yagi, Japan. J. Appl. Phys. 25 (1986)

L333.118. D J. Smith, L. A. Bursill, and D. A. Jefferson, Surf. Sci. 175 (1986) 673.119. D. J. Smith, L. A. Bursill, and M. R. McCartney, Ultramicroscopy 23 (1987) 299.120. M. I. Buckett, J. Strane, D. E. Luzzi, J. P. Zhang, B. W. Wessels, and L. D. Marks, Ultramicroscopy 29

(1989) 217.121. J. L. Hutchison and N. A. Briscoe, Ultramicroscopy 18 (1985) 435.122. Z. C. Kang, D. J. Smith, and L. Eyring, Surf. Sci. 175 (1986) 684.123. X.-G. Ning and H.-Q. Ye, Phil Mag. A62 (1990) 431.124. Z. C. Kang and L. Eyring, Ultramicroscopy 23 (1987) 275.125. A. Andersson, J.-O. Bovin, and P. Walter, J. Catal. 98 (1986) 204.126. J. M. Gibson, M. L. McDonald, and F. C. Unterwald, Phys. Rev. Lett. 55 (1985) 1765.127. P. Lu and D. J. Smith, Surf. Sci. 254 (1991) 119.128. J.-O. Bovin, L. R. Wallenberg, and D. J. Smith, Nature 317 (1985) 47.129. L. R. Wallenberg, J.-O. Bovin, and D. J. Smith, Naturwiss. 72 (1985) 539.130. S. Iijima and T. Ichihashi, Japan. J. Appl. Phys. 24 (1985) 1125.131. L. R. Wallenberg, J.-O. Bovin, A. K. Petford-Long, and D. J. Smith, Ultramicroscopy 20 (1986) 71.132. D. J. Smith, A. K. Petford-Long, L. R. Wallenberg, and J.-O. Bovin, Science 233 (1986) 872.133. A. K. Petford-Long, N. J. Long, D. J. Smith, L. R. Wallenberg, and J.-O. Bovin, In: Physics and

Chemistry of Small Clusters, P. Jena, B. K. Rao, and S. N. Khanna (Eds.), Plenum, New York (1987)p. 127.

134. J.-O. Bovin, J.-O. Malm, A. K. Petford-Long, D. J. Smith, and G. Schmid, Z. Angew. Chem. 27 (1988)555.


135. P. M. Ajayan and L. D. Marks, Phys. Rev. Lett. 60 (1988) 585.136. K.-H. Herrmann, D. Krahl, H.-P. Rust and O. Ulrichs, Optik 44 (1976) 393.137. O. L. Krivanek and P. E. Mooney, Ultramicroscopy 49 (1993) 95.138. W. J. de Ruijter and J. K. Weiss , Rev. Sci. Instr. 63 (1992) 4314.139. W. J. de Ruijter, Micron 26 (1995) 247.140. W. J. de Ruijter and J. K. Weiss, Ultramicroscopy 50 (1993) 269.141. W. O. Saxton, D. J. Smith, and S. J. Erasmus, J. Micros. 130 (1983) 187.142. O. L. Krivanek and M. L. Leber, In: Proc. 51st. Ann. Meet. Microscope Society of America, G. W. Bailey

and C. L. Rieder (Eds.), San Francisco Press, San Francisco (1993) p. 972.143. W. O. Saxton, Ultramicroscopy 58 (1995) 239.144. O. L. Krivanek and P. A. Stadelmann, Ultramicroscopy 60 (1995) 103.145. O. Scherzer, J. Appl. Phys. 20 (1949) 20.146. H. Rose, Ultramicroscopy 56 (1994) 11.147. M. Haider, H. Rose, S. Uhlemann, E. Schwan, B. Kabius, and K. Urban, Ultramicroscopy 75 (1998)

53.148. O. L. Krivanek, N. Dellby, and A. R. Lupini, Ultramicroscopy 78 (1999) 1.149. M. Haider and J. Zach, In: Proc. Microscopy and Microanalysis 1995, G. W. Bailey, M. H. Ellisman,

R. A. Henninger and N. Zaluzec (Eds.), Jones and Begall, New York (1995) p. 596.150. H. Rose, Optik 85 (1990) 95.151. H. W. Mook and P. Kruit, Ultramicroscopy 81 (2000) 129.152. M. H. F. Overwijk, A. J. Bleeker, and A. Thust, Ultramicroscopy 67 (1997) 163.153. G. Mobus, R. Schweinfest, T. Gemming, T. Wagner, and M. Ruhle, J. Microscopy 190 (1998) 109.154. T. Hoche, P. R. Kenway, H.-J. Kleebe, and M. Ruhle, In: Atomic Scale Imaging of Surfaces and Interfaces,

D. Biegelsen, D. J. Smith, and S.-Y. Tong (Eds.), MRS Symp. Proc. Vol. 295, Materials ResearchSociety, Pittsburgh (1993) p. 115.

155. W. E. King and B. S. Lamver, In: Microbeam Analysis, D. G. Howitt (Ed.), San Francisco Press, SanFrancisco (1991) p. 217.

156. W. E. King and G. H. Campbell, Ultramicroscopy 51 (1993) 128.157. K. Urban, B. Kabius, M. Haider, and H. Rose, J. Electron Microscopy, 48 (1999) 821.158. J. H. Chen, H. W. Zandbergen, and D. van Dyck, Ultramicroscopy (2004) in press.159. M. A. O’Keefe, Microsc. & Microanal. 4 (1998) 382.160. C. L. Jia, M. Lentzen, and K. Urban, Science 299 (2003) 870.161. M. J. Hytch and W. M. Stobbs, Ultramicroscopy 53 (1994) 191.162. C. B. Boothroyd, J. Microscopy 190 (1998) 99.163. C. B. Boothroyd, Ultramicroscopy 83 (2000) 159.164. A. Howie, Ultramicrosopy (2004) in press.165. C. B. Boothroyd and R. E. Dunin-Borkowski, Ultramicroscopy (2004) in press.166. L. W. Hobbs, In: Quantitative Electron Microscopy, J. N. Chapman and A. J. Craven (Eds.), Scottish

Universities Summer School in Physics, Edinburgh (1984) p. 399.167. G. Dehm, K. Nadarzinski, F. Ernst, and M. Ruhle, Ultramicroscopy 63 (1996) 49.168. D. Hofmann and F. Ernst, Ultramicroscopy 53 (1994) 205.169. M. J. Beeching and A. E. C. Spargo, Ultramicroscopy 52 (1993) 243.170. M. J. Beeching and A. E. C. Spargo, J. Microscopy 190 (1998) 262.171. M. Lentzen M and K. Urban, Ultramicroscopy 62 (1996) 89.

Documents

14. HIGH RESOLUTION TRANSMISSION ELECTRON MICROSCOPYmx.nthu.edu.tw/~zhwei/member/handbook/Ch14_427-453.pdf · 2008-07-07 · 14. High Resolution Transmission Electron Microscopy 429