Search

The ISTEM mode for TEM has been demonstrated to have several advantages in regard to resolution and precision. While previous works primarily focussed on the advantages due to the reduced spatial coherence, the actual image contrast, i.e. how bright or dark certain atom columns are imaged, has mostly been of secondary concern. The present work sets out to achieve the contrast of annular bright field STEM in ISTEM, producing the high contrast of light elements, for which this method is popular. It is shown from theoretical considerations that using an annular condenser aperture this aim can be realised. The optimal size of this aperture is found by simulative studies. It is then manufactured from platinum foil and installed in an image-aberration corrected microscope. ABF-like ISTEM images of strontium titanate in [100] projection are acquired. The pure oxygen columns are clearly resolved with significant contrast. The image pattern is indeed identical to what is achieved by ABF STEM. A close look at the image formation also shows that the dose needed for a given signal-to-noise ratio is at least a quarter smaller for ABF-like ISTEM compared to ABF STEM, assuming detectors of similar detective quantum efficiency.

In this paper we perform angular resolved annular-dark field (ADF) scanning-transmission electron microscopy (STEM) to study the scattered intensity in an InGaN layer buried in GaN as a function of the scattering angle. We achieved angular resolution with a motorized iris aperture in front of the ADF detector. Using this setup, we investigated how the intensities measured in various angular ranges agree with multislice simulations in the frozen-lattice approximation. We observed a strong influence of relaxation induced surface-strain fields on the ADF intensity, measured its angular characteristics and compared the result with simulations. To assess the agreement of the measured intensity with simulations, we evaluated the specimen thickness in GaN and the indium concentration in InGaN for each angular interval by comparing the measured intensities with simulations. The thickness was strongly overestimated for scattering angles below 40mrad and also the evaluated indium concentration varies with the considered angular range. Using simulations, we investigated which angular ranges show a high sensitivity to variations of the thickness and which intervals strongly depend on the indium concentration. By combining two angular intervals, the indium concentration and the specimen thickness were determined simultaneously, which has potential advantages over the usual quantification method. It is shown that inelastic scattering, surface contamination and mistilt can have an influence on the measured intensity, especially at lower scattering angles below 30-50mrad, which might explain the observed difference between the frozen lattice simulation and the experiment.

The measurement of electric fields in scanning transmission electron microscopy (STEM) is a highly investigated field of research. The constant improvement of spatial resolution in STEM and the development of new hardware for the fast acquisition of diffraction patterns even paved the way for the measurement of atomic electric fields. Although the basic principle that an electric field leads to a tilt of the focussed electron probe that can be detected as a shift of the diffraction pattern in the back focal plane of the objective lens seems quite simple, many challenges arose in the measurement of fields in a quantitative way. In the present study we investigate whether a shift of the diffraction pattern that occurs at an interface between two materials can be related to the electric field which is caused by the difference of the mean inner potentials of the two materials. To this end, experiments and simulations are compared. It is demonstrated that the difference in mean inner potential has an influence on the observed effect, but a quantitative interpretation is difficult. The influence of image recording effects such as shot noise and the modulation transfer function are investigated as well as further effects such as e.g. sample tilt. In addition, the influence of the observed effect on a strain measurement is shown.

In this paper we study the effect of lens aberrations (spherical aberration and astigmatism), beam tilt, contamination and shot noise on the accuracy and precision of position determination in imaging scanning transmission electron microscopy (ISTEM) on the example of BaTiO

4D-scanning transmission electron microscopy (4D-STEM) can be used to measure electric fields such as atomic fields or polarization-induced electric fields in crystal heterostructures. The paper focuses on effects occurring in 4D-STEM at interfaces, where two model systems are used: an AlN/GaN nanowire superlattice as well as a GaN/vacuum interface. Two different methods are applied: First, we employ the centre-of mass (COM) technique which uses the average momentum transfer evaluated from the intensity distribution in the diffraction pattern. Second, we measure the shift of the undiffracted disc (disc-detection method) in nano-beam electron diffraction (NBED). Both methods are applied to experimental and simulated 4D-STEM data sets. We find for both techniques distinct variations in the momentum transfer at interfaces between materials: In both model systems, peaks occur at the interfaces and we investigate possible sources and routes of interpretation. In case of the AlN/GaN superlattice, the COM and disc-detection methods are used to measure internal polarization-induced electric fields and we observed a reduction of the measured fields with increasing specimen thickness.

Strain analyses from experimental series of nano-beam electron diffraction (NBED) patterns in scanning transmission electron microscopy are performed for different specimen tilts. Simulations of NBED series are presented for which strain analysis gives results that are in accordance with experiment. This consequently allows to study the relation between measured strain and actual underlying strain. A two-tilt method which can be seen as lowest-order electron beam precession is suggested and experimentally implemented. Strain determination from NBED series with increasing beam convergence is performed in combination with the experimental realization of a probe-forming aperture with a cross inside. It is shown that using standard evaluation techniques, the influence of beam convergence on spatial resolution is lower than the influence of sharp rings around the diffraction disc which occur at interfaces and which are caused by the tails of the intensity distribution of the electron probe. (C) 2018 Elsevier B.V. All rights reserved.

In this paper we study the effect of lens aberrations (spherical aberration and astigmatism), beam tilt, contamination and shot noise on the accuracy and precision of position determination in imaging scanning transmission electron microscopy (ISTEM) on the example of BaTiO 3 . ISTEM images are simulated as a function of sample thickness and defocus starting from a nearly perfect microscope setting. A defocus range was identified, in which atom column positions were reliably visible and could be decently measured. By averaging over this defocus and thickness range a figure of merit was defined and used to study the influence of above mentioned disturbing effects as a function of their strength. It turned out that column positions might become inaccurate, but distances are measured accurately. These were used to obtain recommendations for the experimental setup to measure the atomic arrangement that induces ferroelectric switching of BaTiO 3 .

This study addresses the comparison of scanning transmission electron microscopy (STEM) measurements of momentum transfers using the first moment approach and the established method that uses segmented annular detectors. Using an ultrafast pixelated detector to acquire four-dimensional, momentum-resolved STEM signals, both the first moment calculation and the calculation of the differential phase contrast (DPC) signals are done for the same experimental data. In particular, we investigate the ability to correct the segment-based signal to yield a suitable approximation of the first moment for cases beyond the weak phase object approximation. It is found that the measurement of momentum transfers using segmented detectors can approach the first moment measurement as close as 0.13 h/nm in terms of a root mean square (rms) difference in 10 nm thick SrTiO3 for a detector with 16 segments. This amounts to 35% of the rms of the momentum transfers. In addition, we present a statistical analysis of the precision of first moment STEM as a function of dose. For typical experimental settings with recent hardware such as a Medipix3 Merlin camera attached to a probe-corrected STEM, we find that the precision of the measurement of momentum transfers stagnates above certain doses. This means that other instabilities such as specimen drift or scan noise have to be taken into account seriously for measurements that target, e.g., the detection of bonding effects in the charge density.

Strain analysis by nano-beam electron diffraction allows for measurements of strain with nanometre resolution in a large field of view. This is done by evaluating distances between diffraction discs in diffraction patterns acquired while a focussed electron beam is scanned across the sample in a transmission electron microscope. The bottleneck of this method is a precise determination of diffraction disc positions, which suffers from the inner structure of the discs caused by dynamical diffraction. Electron beam precession is a tool that solves this problem but it is not commonly available in every microscope. Without precession significant progress has been reported recently by using patterned condenser apertures. The pattern of the aperture is reproduced in patterns of the diffraction discs allowing for a more precise position determination. In this report the accuracy of measured strain profiles using patterned apertures is investigated by evaluation of realistic simulations. This is done especially at interfaces between regions with different lattice plane spacing. It is found by evaluation of the simulations that measured strain profiles are more blurred and hence the accuracy at the interface is worse the more patterns are imprinted to the condenser aperture. An explanation of this effect is given and as a proof of principle a solution to this problem is provided applying geometric phase analysis ptychography.

Several instrumental imperfections of transmission electron microscopes are characterized and their effects on the results of quantitative scanning electron microscopy (STEM) are investigated and quantified using simulations. Methods to either avoid influences of these imperfections during acquisition or to include them in reference calculations are proposed. Particularly, distortions inflicted on the diffraction pattern by an image-aberration corrector can cause severe errors of more than 20% if not accounted for. A procedure for their measurement is proposed here. Furthermore, afterglow phenomena and nonlinear behavior of the detector itself can lead to incorrect normalization of measured intensities. Single electrons accidentally impinging on the detector are another source of error but can also be exploited for threshold-less calibration of STEM images to absolute dose, incident beam current determination and measurement of the detector sensitivity.

The determination of atom positions from atomically resolved transmission electron micrographs is fundamental for the analysis of crystal defects and strain. In recent years annular bright-field (ABF) imaging has become a popular imaging technique owing to its ability to map both light and heavy elements. Contrast formation in ABF is partially governed by the phase of the electron wave, which renders the technique more sensitive to the tilt of the electron beam with respect to the crystal zone axis than high-angle annular dark-field imaging. Here we show this sensitivity experimentally and use image simulations to quantify this effect. This is essential for error estimation in future quantitative ABF studies.

Measurement of lattice strain is important to characterize semiconductor nanostructures. As strain has large influence on the electronic band structure, methods for the measurement of strain with high precision, accuracy and spatial resolution in a large field of view are mandatory. In this paper we present a theoretical study of precision and accuracy of measurement of strain by convergent nano-beam electron diffraction. It is found that the accuracy of the evaluation suffers from halos in the diffraction pattern caused by a variation of strain within the area covered by the focussed electron beam. This effect, which is expected to be strong at sharp interfaces between materials with different lattice plane distances, will be discussed for convergent-beam electron diffraction patterns using a conventional probe and for patterns formed by a precessing electron beam. Furthermore, we discuss approaches to optimize the accuracy of strain measured at interfaces. The study is based on the evaluation of diffraction patterns simulated for different realistic structures that have been investigated experimentally in former publications. These simulations account for thermal diffuse scattering using the frozen-lattice approach and the modulation-transfer function of the image-recording system. The influence of Poisson noise is also investigated.

Images acquired in transmission electron microscopes can be distorted for various reasons such as e.g. aberrations of the lenses of the imaging system or inaccuracies of the image recording system. This results in inaccuracies of measures obtained from the distorted images. Here we report on measurement and correction of elliptical distortions of diffraction patterns. The effect of this correction on the measurement of crystal lattice strain is investigated. We show that the effect of the distortions is smaller than the precision of the measurement in cases where the strain is obtained from shifts of diffracted discs with respect to their positions in images acquired in an unstrained reference area of the sample. This can be explained by the fact that diffraction patterns acquired in the strain free reference area of the sample are distorted in the same manner as the diffraction patterns acquired in the strained region of interest. In contrast, for samples without a strain free reference region such as nanoparticles or nanoporous structures, where we evaluate ratios of lattice plane distances along different directions, the distortions are usually not negligible. Furthermore, two techniques for the detection of diffraction disc positions are compared showing that for samples in which the crystal orientation changes over the investigated area it is more precise to detect the positions of many diffraction discs simultaneously instead of detecting each disc position independently.

For simulation of transmission electron microscopic images and diffraction patterns, the accurate inclusion of thermal diffuse scattering by phonons is important. In the frozen phonon multislice algorithm, this is possible, if thermal displacements according to the realistic, quantum mechanical distribution can be generated. For pure crystals, quantum mechanical calculations based on DFT yield those displacements. But for alloys one is usually restricted to the Einstein approximation, where correlations between atoms are neglected. In this article, molecular dynamics simulations are discussed and used as an alternative method for displacement calculation. Employing an empirical Stillinger-Weber type potential, classical motion is used as an approximation for the quantum mechanical dynamics. Thereby, correlations and possible static atomic displacements are inherently included. An appropriate potential is devised for AlGaN by fitting to force constant matrices determined from DFT and elastic constants of AlN and GaN. A comparison shows that the empiric potential reproduces phonon dispersions and displacement expectations from DFT references. The validity for alloys is successfully demonstrated by comparison to DFT calculations in special quasirandom structures. Subsequently, molecular dynamics were used in multislice simulations of both conventional and scanning TEM images. The resulting images are in very good agreement with DFT based calculations, while a slight yet significant deviation from Einstein approximation results can be seen, which can be attributed to the neglect of correlations in the latter. The presented potential hence proves to be a useful tool for accurate TEM simulations of AlGaN alloys.

The chemical composition of four Si1-xGex layers grown on silicon was determined from quantitative scanning transmission electron microscopy (STEM). The chemical analysis was performed by a comparison of the high-angle annular dark field (HAADF) intensity with multislice simulations. It could be shown that amorphous surface layers originating from the preparation process by focused-ion beam (FIB) at 30 kV have a strong influence on the quantification: the local specimen thickness is overestimated by approximately a factor of two, and the germanium concentration is substantially underestimated. By means of simulations, the effect of amorphous surface layers on the HAADF intensity of crystalline silicon and germanium is investigated. Based on these simulations, a method is developed to analyze the experimental HAADF-STEM images by taking the influence of the amorphous layers into account which is done by a reduction of the intensities by multiplication with a constant factor. This suggested modified HAADF analysis gives germanium concentrations which are in agreement with the nominal values. The same TEM lamella was treated with low-voltage ion milling which removed the amorphous surface layers completely. The results from subsequent quantitative HAADF analyses are in agreement with the nominal concentrations which validates the applicability of the used frozen-lattice based multislice simulations to describe the HAADF scattering of Si1-xGex in STEM. (C) 2017 Elsevier B.V. All rights reserved.

This paper addresses a novel approach to atomic resolution elemental mapping, demonstrating a method that produces elemental maps with a similar resolution to the established method of electron energy-loss spectroscopy in scanning transmission electron microscopy. Dubbed energy-filtered imaging scanning transmission electron microscopy (EFISTEM) this mode of imaging is, by the quantum mechanical principle of reciprocity, equivalent to tilting the probe in energy-filtered transmission electron microscopy (EFTEM) through a cone and incoherently averaging the results. In this paper we present a proof-of-principle EFISTEM experimental study on strontium titanate. The present approach, made possible by chromatic aberration correction, has the advantage that it provides elemental maps which are immune to spatial incoherence in the electron source, coherent aberrations in the probe-forming lens and probe jitter. The veracity of the experiment is supported by quantum mechanical image simulations, which provide an insight into the image-forming process. Elemental maps obtained in EFTEM suffer from the effect known as preservation of elastic contrast, which, for example, can lead to a given atomic species appearing to be in atomic columns where it is not to be found. EFISTEM very substantially reduces the preservation of elastic contrast and yields images which show stability of contrast with changing thickness. The experimental application is demonstrated in a proof-of-principle study on strontium titanate.

A relatively simple yet accurate analytical model for the image formation in the imaging scanning TEM (ISTEM) imaging mode, which implements partial spatial incoherence using a combination of scanning illumination and conventional imaging, is presented. Based on an object function approximation the ISTEM intensity can be divided into a constant, a linear and a nonlinear term. Under certain conditions, which are discussed, the formation of both linear and nonlinear terms can be expressed by convolutions with point spread functions. A closer inspection of these allows an insight into the advantages of ISTEM compared to conventional TEM (CTEM). The findings of the proposed model are confirmed by comparison to multislice simulations. A close investigation of the linear coherent contrast transfer function allows the derivation of optimal imaging conditions to reach a maximum resolution for a given signal-to-noise ratio. The robustness of ISTEM towards temporal incoherence is finally demonstrated and discussed within the model. (C) 2017 Elsevier B.V. All rights reserved.

In this paper, we investigate how precise atoms of a small nanocluster can ultimately be located in three dimensions (3D) from a tilt series of images acquired using annular dark field (ADF) scanning transmission electron microscopy (STEM). Therefore, we derive an expression for the statistical precision with which the 3D atomic position coordinates can be estimated in a quantitative analysis. Evaluating this statistical precision as a function of the microscope settings also allows us to derive the optimal experimental design. In this manner, the optimal angular tilt range, required electron dose, optimal detector angles, and number of projection images can be determined.

Nano-beam electron diffraction (NBED) is a method which can be applied to measure lattice strain and polarisation fields in strained layer heterostructures and transistors. To investigate precision, accuracy and spatial resolution of such measurements in dependence of properties of the specimen as well as electron optical parameters, simulations of NBED patterns are required which allow to predict the result of common disc-detection algorithms. In this paper we demonstrate by focusing on the detection of the central disc in crystalline silicon that such simulations require to take several experimental characteristics into account in order to obtain results which are comparable to those from experimental NBED patterns. These experimental characteristics are the background intensity, the presence of Poisson noise caused by electron statistics and blurring caused by inelastic scattering and by the transfer quality of the microscope camera. By means of these optimized simulations, different effects of specimen properties on disc detection - such as strain, surface morphology and compositional changes on the nanometer scale - are investigated and discussed in the context of misinterpretation in experimental NBED evaluations. It is shown that changes in surface morphology and chemical composition lead to measured shifts of the central disc in the NBED pattern of tens to hundreds of grad. These shifts are of the same order of magnitude or even larger than shifts that could be caused by an electric polarisation field in the range of MV/cm. (C) 2017 Elsevier B.V. All rights reserved.

Recently, imaging scanning transmission electron microscopy (ISTEM) has been proposed as a promising new technique combining the advantages of conventional TEM (CTEM) and STEM (Rosenauer et al., 2014 [1]). The ability to visualize light and heavy elements together makes it a particularly interesting new, spatially incoherent imaging mode. Here, we evaluate this technique in term of precision with which atomic column locations can be measured. By using statistical parameter estimation theory, we will show that these locations can be accurately measured with a precision in the picometer range. Furthermore, a quantitative comparison is made with HAADF STEM imaging to investigate the advantages of ISTEM.

This study sheds light on the prerequisites, possibilities, limitations and interpretation of high-resolution differential phase contrast (DPC) imaging in scanning transmission electron microscopy (STEM). We draw particular attention to the well-established DPC technique based on segmented annular detectors and its relation to recent developments based on pixelated detectors. These employ the expectation value of the momentum transfer as a reliable measure of the angular deflection of the STEM beam induced by an electric field in the specimen. The influence of scattering and propagation of electrons within the specimen is initially discussed separately and then treated in terms of a two-state channeling theory. A detailed simulation study of GaN is presented as a function of specimen thickness and bonding. It is found that bonding effects are rather detectable implicitly, e.g., by characteristics of the momentum flux in areas between the atoms than by directly mapping electric fields and charge densities. For strontium titanate, experimental charge densities are compared with simulations and discussed with respect to experimental artifacts such as scan noise. Finally, we consider practical issues such as figures of merit for spatial and momentum resolution, minimum electron dose, and the mapping of larger-scale, built-in electric fields by virtue of data averaged over a crystal unit cell. We find that the latter is possible for crystals with an inversion center. Concerning the optimal detector design, this study indicates that a sampling of 5 mrad per pixel is sufficient in typical applications, corresponding to approximately 10 × 10 available pixels.

We investigated segregation of indium in an In x Ga 1− x As/GaAs heterostructure via high-angle annular dark field-scanning transmission electron microscopy (HAADF-STEM), where contrast strongly depends on the nuclear charges of the scattering atoms (Z-contrast). Indium concentration maps have been deduced from HAADF-STEM images by comparing normalized measured intensities with multislice simulations in the frozen lattice approach. Segregation coefficients were derived following the segregation model of Muraki et al. [1] . This is demonstrated for HAADF-STEM images recorded in [100] and [110] zone-axes. Determined indium concentrations and segregation coefficients are compared with results from composition analysis by lattice fringe analysis (CELFA) measurements and energy-dispersive X-ray analysis (EDX).

The nitrogen concentration of GaN 0.01 ≤ x ≤ 0.05 As 1 − x quantum wells was determined from high resolution scanning transmission electron microscopy (HRSTEM) images taken with a high-angle annular dark field (HAADF) detector. This was done by applying two independent methods: evaluation of the scattering intensity and strain state analysis. The HAADF scattering intensity was computed by multislice simulations taking into account the effect of static atomic displacements and thermal diffuse scattering. A comparison of the mean intensity per atom column on the experimental images with these simulations enabled us to generate composition maps with atomic scale resolution. STEM simulations of large supercells proved that local drops of the HAADF intensity observed close to embedded quantum wells are caused by surface strain relaxation. The same STEM images were evaluated by strain state analysis. We suggest a real space method which is not affected by fly-back errors in HRSTEM images. The results of both evaluation methods are in accordance with data obtained from X-ray diffraction measurements.

In scanning transmission electron microscopy using a high-angle annular dark field detector, image intensity strongly depends on specimen thickness and composition. In this paper we show that measurement of image intensities relative to the intensity of the incoming electron beam allows direct comparison with simulated image intensities, and thus quantitative measurement of specimen thickness and composition. Simulations were carried out with the frozen lattice and absorptive potential multislice methods. The radial inhomogeneity of the detector was measured and taken into account. Using a focused ion beam (FIB) prepared specimen we first demonstrate that specimen thicknesses obtained in this way are in very good agreement with a direct measurement of the thickness of the lamella by scanning electron microscopy in the FIB. In the second step we apply this method to evaluate the composition of Al x Ga 1 - x N / GaN layers. We measured ratios of image intensities obtained in regions with unknown and with known Al-concentration x , respectively. We show that estimation of the specimen thickness combined with evaluation of intensity ratios allows quantitative measurement of the composition x . In high-resolution images we find that the image intensity is well described by simulation if the simulated image is convoluted with a Gaussian with a half-width at half-maximum of 0.07 nm.

We present a new method to measure structure factors from electron spot diffraction patterns recorded under almost parallel illumination in transmission electron microscopes. Bloch wave refinement routines have been developed to refine the crystal thickness, its orientation and structure factors by comparison of experimentally recorded and calculated intensities. Our method requires a modicum of computational effort, making it suitable for contemporary personal computers. Frozen lattice and Bloch wave simulations of GaAs diffraction patterns are used to derive optimised experimental conditions. Systematic errors are estimated from the application of the method to simulated diffraction patterns and rules for the recognition of physically reasonable initial refinement conditions are derived. The method is applied to the measurement of the 200 structure factor for GaAs. We found that the influence of inelastically scattered electrons is negligible. Additionally, we measured the 200 structure factor from zero loss filtered two-dimensional convergent beam electron diffraction patterns. The precision of both methods is found to be comparable and the results agree well with each other. A deviation of more than 20% from isolated atom scattering data is observed, whereas close agreement is found with structure factors obtained from density functional theory [A. Rosenauer, M. Schowalter, F. Glas, D. Lamoen, Phys. Rev. B 72 (2005), 085326-1], which account for the redistribution of electrons due to chemical bonding via modified atomic scattering amplitudes.

Inelastic image simulation software is presented, implementing the double channeling approximation which takes into account the combination of multiple elastic and single inelastic scattering in a crystal. The approach is described with a density matrix formalism. Two applications in high resolution energy filtered (EFTEM) transmission electron microscopy (TEM) images are presented: thickness-defocus maps for SrTiO 3 and exit plane intensities for an ( LaAlO 3 ) 3 ( SrTiO 3 ) 3 multilayer system. Both systems show a severe breakdown in direct interpretability which becomes worse for higher acceleration voltages, thicker samples and lower excitation edge energies. Since this effect already occurs in the exit plane intensity, it is a fundamental limit and image simulations in EFTEM are indispensable just as they are indispensable for elastic high resolution TEM images.

The electronic properties of quaternary AlInGaN devices significantly depend on the homogeneity of the alloy. The identification of compositional fluctuations or verification of random-alloy distribution is hence of grave importance. Here, a comprehensive multiprobe study of composition and compositional homogeneity is presented, investigating AlInGaN layers with indium concentrations ranging from 0 to 17 at% and aluminium concentrations between 0 and 39 at% employing high-angle annular dark field scanning electron microscopy (HAADF STEM), energy dispersive X-ray spectroscopy (EDX) and atom probe tomography (APT). EDX mappings reveal distributions of local concentrations which are in good agreement with random alloy atomic distributions. This was hence investigated with HAADF STEM by comparison with theoretical random alloy expectations using statistical tests. To validate the performance of these tests, HAADF STEM image simulations were carried out for the case of a random-alloy distribution of atoms and for the case of In-rich clusters with nanometer dimensions. The investigated samples, which were grown by metal-organic vapor phase epitaxy (MOVPE), were thereby found to be homogeneous on this nanometer scale. Analysis of reconstructions obtained from APT measurements yielded matching results. Though HAADF STEM only allows for the reduction of possible combinations of indium and aluminium concentrations to the proximity of isolines in the two-dimensional composition space. The observed ranges of composition are in good agreement with the EDX and APT results within the respective precisions.

High angle annular dark field scanning transmission electron microscopy (HAADF STEM) images provide sample information which is sensitive to the chemical composition. The image intensities indeed scale with the mean atomic number Z. To some extent, chemically different atomic column types can therefore be visually distinguished. However, in order to quantify the atomic column composition with high accuracy and precision, model-based methods are necessary. Therefore, an empirical incoherent parametric imaging model can be used of which the unknown parameters are determined using statistical parameter estimation theory (Van Aert et al., 2009, [1]). In this paper, it will be shown how this method can be combined with frozen lattice multislice simulations in order to evolve from a relative toward an absolute quantification of the composition of single atomic columns with mixed atom types. Furthermore, the validity of the model assumptions are explored and discussed.

The present paper deals with an extension of the composition evaluation by lattice fringe analysis (CELFA). The procedure is applied to an InxGa1−xAs/GaAs(0 0 1) Stranski–Krastanow heterostructure capped with 10 nm GaAs. CELFA is based on images obtained with a three-beam condition of only three excited reflections. Their interference produces a fringe pattern in the image plane. Under appropriate imaging conditions, the contrast pattern strongly depends on the In-content. CELFA uses the amplitude ratio |I 0 0 2 |/|I 0 0 4 | of the (0 0 2)- and (0 0 4)-reflections of image cell diffractrograms, which is an approximately linear function of the In-content x. In this paper an extension of CELFA will be introduced that is based on a defocus series instead of only one single image. All free parameters of the image formation are directly derived from the defocus series with a simple procedure. Therefore, the extended CELFA procedure does not require any knowledge concerning the imaging conditions (except the sample orientation). Furthermore, we will show how the effect of locally changing imaging conditions across the HRTEM image can be taken into account.

In this paper we present a compositional analysis, by high-resolution transmission electron microscopy of a 2 nm InxGa1−xAs layer (nominal In-content 60%) which was deposited on GaAs ( 0 0 1 ) and capped with 10 nm GaAs. A three-beam condition close to the (1 0 0) zone axis is used. The lattice fringe images result from the interference of the ( 0 0 0 ), ( 0 0 4 ) and ( 0 0 2 ) beams with the latter centered on the optical axis. Due to the chemical sensitivity of the ( 0 0 2 ) beam, the obtained contrast patterns strongly depend on the In-content x averaged along the electron beam direction. We suggest a simple analysis procedure that uses the ratio A002/A004 of the ( 0 0 2 ) and ( 0 0 4 ) reflections of local image cell diffractograms, respectively, which is a bijective function of x. Furthermore, we show that the suggested method does not require the exact knowledge of the objective lens defocus and the sample thickness in the electron beam direction. In order to check the reliability of the gained results, a finite-element calculation of the strain state is performed which is based on the evaluated In-concentration profile. The results of the finite-element calculation are compared with an evaluation of the image using atomic scale strain measurements. The investigations of the InxGa1−xAs layer showed that the GaAs overgrowth caused a structural modification of the initial island structure which is transformed into an InxGa1−xAs layer with a rather homogeneous thickness and an In-content that varies locally. Reasons for the morphological transition are discussed.

Thin-film-based phase-plates are applied to enhance the contrast of weak-phase objects in transmission electron microscopy. In this work, metal-film-based phase-plates are considered to reduce contamination and electrostatic charging, which up to now limit the application of phase-plates fabricated from amorphous C-films. Their crystalline structure requires a model for the simulation of the effect of crystallinity on the phase-plate properties and the image formation process. The model established in this work is verified by experimental results obtained by the application of a textured nanocrystalline Au-film-based Hilbert phase-plate. Based on the model, it is shown that monocrystalline and textured nanocrystalline phase-plate microstructures of appropriate thickness and crystalline orientation can be a promising approach for phase-contrast transmission electron microscopy.

The Stobbs factor problem, a major difference of absolute contrast between experimental and simulated high resolution transmission electron microscopy images has been reported frequently. In this respect, some modifications to the multislice simulation techniques were proposed to improve the correspondence to the experiment. The influence of different suggestions on the simulated contrast is investigated by numerical simulations. The agreement between experiment and simulations is then checked by comparison with high-resolution micrographs of crystalline gold. The experimental data is therefore compared to simulated intensities computed for the thickness and orientation of the specimen measured by refinements of diffraction patterns. The agreement of both intensity and contrast is investigated and the remaining contrast discrepancy is determined. The results show a good agreement for a small objective aperture, while for a larger aperture a difference of contrast by a factor of 1.2 can still be observed. Without any aperture, the deviation between experiment and simulations is largest.

In the present paper, a statistical model-based method to count the number of atoms of monotype crystalline nanostructures from high resolution high-angle annular dark-field (HAADF) scanning transmission electron microscopy (STEM) images is discussed in detail together with a thorough study on the possibilities and inherent limitations. In order to count the number of atoms, it is assumed that the total scattered intensity scales with the number of atoms per atom column. These intensities are quantitatively determined using model-based statistical parameter estimation theory. The distribution describing the probability that intensity values are generated by atomic columns containing a specific number of atoms is inferred on the basis of the experimental scattered intensities. Finally, the number of atoms per atom column is quantified using this estimated probability distribution. The number of atom columns available in the observed STEM image, the number of components in the estimated probability distribution, the width of the components of the probability distribution, and the typical shape of a criterion to assess the number of components in the probability distribution directly affect the accuracy and precision with which the number of atoms in a particular atom column can be estimated. It is shown that single atom sensitivity is feasible taking the latter aspects into consideration.

Single-crystalline ZnTe epilayers were grown by atmospheric-pressure metal organic vapor phase epitaxy (MOVPE) and hot wall epitaxy (HWE) on (001) GaAs. The misfit of lattice constants is -7.5% at the growth temperature of 350°C. Conventional transmission electron microscopy (TEM) and high-resolution transmission electron microscopy (HRTEM) have been used to investigate the misfit dislocations in the ZnTe/GaAs (001) system. The types and distribution of the observed dislocations are independent of growth techniques used. The most common line defects at the ZnTe/GaAs interface are 60° and Lomer dislocations with Burgers vectors of 1 2 a〈110〉. Their respective abundance ratio is 2:1. Less than 3% of the 60° dislocations dissociate into partial dislocations limiting a stacking fault. The distances between parallel {111} planes terminating at the interface have been determined from an analysis of about 1 μm of the projected ZnTe/GaAs interface. The distribution fits by a Gaussian with an average distance of 57 A. This allows an estimate of the residual biaxial strain at the interface which is found to be compressive with a magnitude less than 0.6%. The distances of the observed dislocation cores form a histogram. Its special shape shows four maxima at about 0, 5, 9 and 12× 1 2 as〈110〉. They can be explained by a model for the correlated nucleation of misfit dislocations in highly mismatched heterostructures like ZnTe/GaAs.

Thermal diffuse scattered electrons significantly contribute to high-resolution transmission electron microscopy images. Their intensity adds to the background and is peaked at positions of atomic columns. In this paper we suggest an approximation to simulate intensity of thermal diffuse scattered electrons in plane-wave illumination transmission electron microscopy using an emission-potential multislice algorithm which is computationally less intensive than the frozen lattice approximation or the mutual intensity approach. Intensity patterns are computed for Au and InSb for different crystal orientations. These results are compared with intensities from the frozen lattice approximation based on uncorrelated vibration of atoms as well as with the frozen phonon approximation for Au. The frozen phonon method uses a detailed phonon model based on force constants we computed by a density functional theory approach. The comparison shows that our suggested emission-potential method is in close agreement with both the frozen lattice and the frozen phonon approximations.

The mean inner potentials of various III-V semiconductors, Si and Ge have been calculated by density functional theory methods. For that purpose, the Coulomb potential of slabs consisting of a crystal and vacuum region has been computed and averaged inside the crystal region. The computed values are in agreement with experimental values obtained by electron holography for Si and GaAs. For the other semiconductors, the deviations are smaller than 0.8 V. The results from density functional theory are approximately 10% smaller than the values derived from atomic scattering factors computed by Hartree Fock calculations.

The mean inner potential of GaAs(14.18V), InAs(14.50V), GaP(14.35V) and InP(14.50V) has been measured by transmission electron holography using the phase shift of the (000)-beam of the first hologram sideband. To provide a defined specimen geometry we used 90 degrees wedges obtained by the cleavage technique. The exact excitation condition as well as the acceleration voltage of the electrons were determined from convergent beam electron diffraction images. The magnification is extracted from two-beam lattice fringe images and dynamical effects are taken into account by Bloch-wave calculations.

A method for compositional analysis of low-dimensional heterostructures is presented. The suggested procedure is based on electron holography and the exploitation of the chemically sensitive (0 0 2) reflection. We apply an off-axis imaging condition with the (0 0 2) beam strongly excited and centered on the optic axis. The first side band of the hologram is centered using an "empty" reference hologram obtained for a hole of the specimen. From the centered side band we use the phase of the central (0 0 0) and the amplitude of the (0 0 2) reflections to evaluate the local composition and the local specimen thickness in an iterative and self-consistent way. Delocalization effects that lead to a shift of the spatial information of (0 0 0) and (0 0 2) reflections are taken into account. The application of the procedure is demonstrated with an AlAs/GaAs(0 0 1) superlattice with a period of 5 nm. The concentration profiles obtained are discussed in relation to segregation. The measured segregation efficiency is R = 0.51 +/- 0.02. (C) 2001 Elsevier Science B.V. All rights reserved.