ADF Imaging Research Articles

Trainable segmentation for transmission electron microscope images of inorganic nanoparticles.

We present a trainable segmentation method implemented within the python package ParticleSpy. The method takes user labelled pixels, which are used to train a classifier and segment images of inorganic nanoparticles from transmission electron microscope images. This implementation is based on the trainable Waikato Environment for Knowledge Analysis (WEKA) segmentation, but is written in python, allowing a large degree of flexibility and meaning it can be easily expanded using other python packages. We find that trainable segmentation offers better accuracy than global or local thresholding methods and requires as few as 100 user-labelled pixels to produce an accurate segmentation. Trainable segmentation presents a balance of accuracy and training time between global/local thresholding and neural networks, when used on transmission electron microscope images of nanoparticles. We also quantitatively investigate the effectiveness of the components of trainable segmentation, its filter kernels and classifiers, in order to demonstrate the use cases for the different filter kernels in ParticleSpy and the most accurate classifiers for different data types. A set of filter kernels is identified that are effective in distinguishing particles from background but that retain dissimilar features. In terms of classifiers, we find that different classifiers perform optimally for different image contrast; specifically, a random forest classifier performs best for high-contrast ADF images, but that QDA and Gaussian Naïve Bayes classifiers perform better for low-contrast TEM images.

Atomic-scale investigation of the heterogeneous structure and ionic distribution in an ionic liquid using scanning transmission electron microscopy.

Ionic liquids show characteristic properties derived from them being composed of only molecular ions, and have recently been used as solvents for chemical reactions and as electrolytes for electrochemical devices. The liquid structures, i.e., ionic distributions, form when solutes are dissolved in ionic liquids and fundamentally affect the reactions and transfer efficiency in such solutions. In this study, we directly observe the liquid structure in a solution of the long-chain ionic liquid 1-octyl-3-methylimidazolium bromide (C8mim Br) and barium stearate (Ba(C17H35COO)2) using the annular dark-field method of scanning transmission electron microscopy (ADF-STEM). The ADF image shows a 10 nm-scale heterogeneity in the image intensity, which reflects the heterogeneous ionic distribution in the solution. The number density distributions of all the component ions (C8mim+, Br−, Ba2+, and C17H35COO−) were estimated from the ADF image intensity and then visualized. These ionic distribution maps depicted the spatial relationships between the ions at the sub-nanometer scale and revealed that the heterogeneity is largely derived from the large differences in size, charge distributions, and van der Waals interactions.

Quantitative STEM of Catalyst Nanoparticles using ADF Imaging with Simultaneous EDS and EELS Spectroscopy.

Quantitative STEM of Catalyst Nanoparticles using ADF Imaging with Simultaneous EDS and EELS Spectroscopy.

Quantification of a Heterogeneous Ruthenium Catalyst on Carbon-black using ADF Imaging

Using recent developments in ADF quantification techniques, we utilise a rapid atom counting procedure to document the evolution of a heterogeneous ruthenium catalyst on carbon black. Selected areas of the ruthenium catalyst were imaged for approximately 15 minutes. The imaged areas show that the Ru wets the support forming a thin amorphous layer that transforms to a crystalline layer under beam irradiation.

Is there a Stobbs factor in atomic-resolution STEM-EELS mapping?

Recent work has convincingly argued that the Stobbs factor-disagreement in contrast between simulated and experimental atomic-resolution images-in ADF-STEM imaging can be accounted for by including the incoherent source size in simulation. However, less progress has been made for atomic-resolution STEM-EELS mapping. Here we have performed carefully calibrated EELS mapping experiments of a [101] DyScO3 single-crystal specimen, allowing atomic-resolution EELS signals to be extracted on an absolute scale for a large range of thicknesses. By simultaneously recording the elastic signal, also on an absolute scale, and using it to characterize the source size, sample thickness and inelastic mean free path, we eliminate all free parameters in the simulation of the core-loss signals. Coupled with double channeling simulations that incorporate both core-loss inelastic scattering and dynamical elastic and thermal diffuse scattering, the present work enables a close scrutiny of the scattering physics in the inelastic channel. We found that by taking into account the effective source distribution determined from the ADF images, both the absolute signal and the contrast in atomic-resolution Dy-M5 maps can be closely reproduced by the double-channeling simulations. At lower energy losses, discrepancies are present in the Sc-L2,3 and Dy-N4,5 maps due to the energy-dependent spatial distribution of the background spectrum, core-hole effects, and omitted complexities in the final states. This work has demonstrated the possibility of using quantitative STEM-EELS for element-specific column-by-column atom counting at higher energy losses and for atomic-like final states, and has elucidated several possible improvements for future theoretical work.

Rapid Measurement of Nanoparticle Thickness Profiles

A method to measure the thickness of a single-crystal nanoparticle in the direction parallel to the incident beam from annular dark field scanning transmission electron microscope (ADF-STEM) images is reported, providing a map of thickness versus position across the nanoparticle--a 'thickness profile' image. The method is rapid and hence suitable for surveying large numbers of nanoparticles. The method measures the intensity scattered to a characterised ADF detector and compares this to the incident beam intensity, to obtain a normalized ADF image. The normalised intensity is then converted to thickness via dynamical ADF image simulations. The method is accurate within 10% and the precision is dominated primarily by 'shot noise'. Merits and limitations of this method are discussed. A method to calibrate the response function of the ADF detector without external equipment is also described, which is applicable to the entire range of gain and background settings.

Strain, composition and disorder in ADF imaging of semiconductors

The effect of strain, composition and disorder in ADF images is systematically studied as a function of detection angle in order to understand the main contrast mechanisms. We demonstrate that the complex phenomenology in ADF images can be accounted for by accurate simulations and modelling. The advantage of an accurate modelling on the image interpretation will be demonstrated in the case of dislocations, chemical analysis of InGaP/GaAs and most noticeably the measurement of both In and N content in quaternary InGaAsN.

A new insight on crystalline strain and defect features by STEM–ADF imaging

Scanning TEM with a High Angle Annular Dark Field (HAADF) detector is an outstanding tool for chemical analysis; it permits to image compositional variations at atomic scale as variation in the image intensity. However in the interpretation of HAADF images the role of elastic strain and crystallographic defects are often neglected. For low index specimen orientation crystal imperfections can become very important due to the interplay of strain and channeling effects. Thanks to a recently developed channeling model, the main strain effects on the image can be predicted. An adequate description of these phenomena makes HAADF an interesting technique to complement or substitute traditional TEM characterization in many relevant materials science contexts. Examples will be shown for the case of semiconductor quantum dot, core shell nanowires and dislocations.

The role of rough surfaces in quantitative ADF imaging of gallium nitride-based materials

We have studied the surface of GaN after preparations commonly used to thin cross sectional samples to electron transparency with atomic force microscopy (AFM) and scanning transmission electron microscopy annular dark-field (STEM-ADF) imaging. Tripod polishing and ion milling leave the GaN surface in varying degrees of roughness and strain. By comparing the roughness seen in the AFM with the 'roughness' observed in STEM-ADF imaging, we show most nano-scale variations in ADF can be attributed to surface roughness alone. However, some systematic differences are seen for very thin specimens.

Effects of amorphous layers on ADF-STEM imaging

A study of high-resolution ADF imaging in uncorrected and aberration-corrected STEMs was carried out by multislice simulation. The presence of amorphous layers at the surface of a crystalline specimen is shown to significantly alter the visibility of the atomic columns. After propagating through an amorphous layer a portion of the beam passes without any alteration while scattered electrons introduce a Gaussian background. The dependence of the image contrast on the crystal structure, orientation and the types of the atoms present in the crystal was studied. In the case of uncorrected probes an amorphous layer thicker than 200 Å is necessary to achieve considerable reduction of the visibility of the atomic columns, but with aberration-corrected probes only 60 Å is necessary. With changes in defocus, crystalline specimens with amorphous layers on the top can also be imaged and high-resolution ADF images can be obtained. An amorphous layer at the beam entry surface affects the ADF image more than that of an amorphous layer at the exit surface. Approximately linear reduction of the contrast (with a slop of 1) is expected with increased thickness of amorphous layer.

Au Cluster Behavior Using Aberration Corrected ADF Imaging

Au Cluster Behavior Using Aberration Corrected ADF Imaging

Development of dedicated STEM with high stability

We developed a dedicated scanning transmission electron microscope with high-stability. The mechanical and electronic stabilities of the microscope were substantially improved, e.g. the specimen drift rate was found to be <0.2 nm min(-1). The Fourier transform of an ADF image showed spots of 0.105 nm at an acceleration voltage of 200 kV without spherical aberration corrector. The stabilized STEM instrument allows us to acquire distortion-free STEM images and high-signal to noise ratio analyses. We have shown the outline of the instrument and preliminary results.

Reduction of Contrast in ADF-STEM Images Due To Amorphous Layer

AbstractA study of high-resolution ADF imaging in uncorrected and aberration-corrected STEMs was carried out by multislice simulation. The presence of amorphous layers at the surface of a crystalline specimen is shown to significantly alter the visibility of the atomic columns in ADF images. After propagating through an amorphous layer a portion of the beam passes without any alteration while scattered electrons introduce a Gaussian background. An amorphous layer at the beam entry surface appears to have slightly more of an effect on the ADF image contrast than that of an amorphous layer at the exit surface, and this difference increases with increasing atomic number. With constant crystal layer thickness, the reduction of contrast as a function of increasing amorphous layer is found to have the same behavior, regardless of the thickness of the initial crystal layer thickness.

Inversion domain boundaries in ZnO with additions of Fe 2O 3 studied by high-resolution ADF imaging

Columns of metal atoms in the polytypoid compound Fe 2O 3(ZnO) 15 could be resolved by high angle annular dark field imaging in a transmission electron microscopy (TEM)/STEM electron microscope – a result which could not be realized by high-resolution bright field imaging due to inherent strain from inversion domains and inversion domain boundaries (IDBs) in the crystals. The basal plane IDB was imaged in [ 1 1 ¯ 0 0 ] yielding the spacing of the two adjacent ZnO domains, while imaging in [ 2 1 ¯ 1 ¯ 0 ] yields the position of single metal ions. The images allow the construction of the entire domain structure including the stacking sequence and positions of the oxygen ions. The IDB consists of a single layer of octahedrally co-ordinated Fe 3+ ions, and the inverted ZnO domains are related by point symmetry at the iron position. The FeO 6 octahedrons are compressed along the ZnO c-axis resulting in a Fe O bond length of 0.208 nm which is in the range of Fe O distances in iron containing oxides. The model of the basal plane boundary resembles that of the IDB in polytypoid ZnO–In 2O 3 compounds.

Three-dimensional ADF imaging of individual atoms by through-focal series scanning transmission electron microscopy

Aberration correction in scanning transmission electron microscopy has more than doubled the lateral resolution, greatly improving the visibility of individual impurity or dopant atoms. Depth resolution is increased five-fold, to the nanometer level. We show how a through-focal series of images enables single Hf atoms to be located inside an advanced gate dielectric device structure to a precision of better than 0.1 × 0.1 × 0.5 nm . This depth sectioning method for three-dimensional characterization has potential applications to many other fields, including polycrystalline materials, catalysts and biological structures.

Relationships among equivalent oxide thickness, nanochemistry, and nanostructure in atomic layer chemical-vapor-deposited Hf–O films on Si

The relationships among the equivalent oxide thickness (EOT), nanochemistry, and nanostructure of atomic layer chemical-vapor-deposited (ALCVD) Hf–O-based films, with oxide and nitrided oxide interlayers on Si substrates, were studied using x-ray photoelectron spectroscopy (XPS), high-resolution transmission electron microscopy (HRTEM), scanning transmission electron microscopy (STEM) in annular dark-field imaging (ADF), and parallel electron energy-loss spectroscopy (PEELS), capacitance–voltage, and leakage-current–voltage measurements. The XPS (Hf 4f binding energy shift) studies indicated the formation of Hf–O–Si bonds in as-deposited amorphous films, the amount of which was influenced by the interlayer composition and annealing conditions. After post-deposition annealing in N2 and O2, the Hf–O layers were nanocrystalline. Although HRTEM images showed a structurally sharp interface between the Hf–O layer and the interlayer, angle-resolved XPS, ADF imaging, and PEELS in the STEM revealed a chemically diffused HfSiOx region in between. This interdiffusion was observed by the detection of Si (using Si L edge) and Hf (using Hf O2,3 edge) in the Hf–O layer and the interlayer. For an annealed Hf–O/interlayer stack, with an ALCVD target thickness of 4.0 nm for the Hf–O layer on 1.2 nm of nitrided chemical oxide, the experimentally measured EOT and leakage current (at −1 V) were 1.52 nm and ∼10−8 A/cm2. A three-layer (1.2 nm interlayer of nitrided chemical oxide/compositionally graded, 2 nm region of HfSiOx/2 nm HfO2 layer) capacitor model was used to determine the respective contributions to the measured EOT, and the dielectric permittivity of the interlayer was found to be 6.06. These studies clearly indicate that a total EOT of 1 nm and below is attainable in the Hf–N–O–Si/Si–N–O system.

Contrast Effect of Strain Fields in ADF-STEM Imaging

STEM is an important tool for the study of material by forming Z-contrast ADF images. Studies of strain fields suggest that strain can cause extra contrast in ADF images. Since strain fields exist in many TEM samples, especially interfaces of two different materials, it is necessary to study the contrast effect of strain fields in more detail. in this paper, we study the contrast effect of strain fields by STEM experiments and multislice simulations.Sample thickness and the dimensions of ADF detector can both affect strain contrast. We first study the influence of sample thickness. A cross-section TEM sample of an interface of crystal silicon and amorphous silicon was prepared by typical tripod polishing and was studied with the 100 kV Cornell VG 5.0 UHV STEM. The sample was tilted so that the crystal was on the (110) zone axis and ADF images of the interface were recorded.

Toward Quantitative Annular Dark Field Imaging

Annular Dark Field imaging has the potential to be directly quantifiable. By this I mean that with careful measurement, the absolute image intensity has physical meaning. Unlike Bright Field TEM, the ADF image has no contrast reversals with focus and with the exception of thick specimens there are no contrast reversals with changes in thickness. Thus, image intensity is related to thickness, composition, orientation, and structure of local regions whose size is determined by the electron probe. The ability to extract quantitative information about the specimen from the intensity requires careful collection of the intensity data and a solid understanding of how that intensity will change with thickness, composition, orientation, and structure. The qualitative effect of thickness and composition has been well shown in the literature but more quantitative approaches have been lacking.The simplest models of ADF imaging treat each atom interacting

High Angle Annular Dark Field (HAADF) STEM Tomography of Nanostructured Catalysts

Abstract Progress in the design of catalysts based on nanoparticles supported within a mesoporous silica framework requires structural analysis at high spatial resolution. While bulk analysis by X-ray diffraction and EXAFS can give the structure of active sites they are unable to determine their relative positions and local physical structure. Some success has been achieved using a combination of STEM ADF imaging and EDX mapping to elucidate such structures but these results are limited to giving 2D projections of 3D arrangements. The need exists therefore to analyse specimens in full 3D. A suitable approach is electron tomography where the 3D structure is reconstructed from a tilt series of 2D projections. in conventional electron tomography a tilt series is acquired using bright field (BF) TEM. However the large beam current means such an approach generally cannot be used on specimens based on a mesoporous silica framework as beam damage can develop rapidly.

Atomic resolution EELS analysis of a misfit dislocation at a GeSi/Si interface

A dissociated 60° misfit dislocation at the substrate interface of a Si/GexSi(1−x) heterojunction has been examined using EELS and ADF imaging. New spectra are obtained at the intrinsic stacking fault, at the dislocation cores and in the strained regions on either side of the stacking fault. A splitting of the L1 conduction band due to symmetry breaking at the stacking fault is observed. Near edge conduction band states are verified at the partial dislocation cores, but not at the stacking fault.