Weizong Xu

Quantitative atomic resolution elemental mapping via absolute-scale energy dispersive X-ray spectroscopy.

Quantitative agreement on an absolute scale is demonstrated between experiment and simulation for two-dimensional, atomic-resolution elemental mapping via energy dispersive X-ray spectroscopy. This requires all experimental parameters to be carefully characterized. The agreement is good, but some discrepancies remain. The most likely contributing factors are identified and discussed. Previous predictions that increasing the probe forming aperture helps to suppress the channelling enhancement in the average signal are confirmed experimentally. It is emphasized that simple column-by-column analysis requires a choice of sample thickness that compromises between being thick enough to yield a good signal-to-noise ratio while being thin enough that the overwhelming majority of the EDX signal derives from the column on which the probe is placed, despite strong electron scattering effects.

A numerical model for multiple detector energy dispersive X-ray spectroscopy in the transmission electron microscope.

Here we report a numerical approach to model a four quadrant energy dispersive X-ray spectrometer in the transmission electron microscope. The model includes detector geometries, specimen position and absorption, shadowing by the holder, and filtering by the Be carrier. We show that this comprehensive model accurately predicts absolute counts and intensity ratios as a function of specimen tilt and position. We directly compare the model to experimental results acquired with a FEI Super-X EDS four quadrant detector. The contribution from each detector to the sum is investigated. The program and source code can be downloaded from https://github.com/subangstrom/superAngle.

Structure and chemistry of passivated SiC/SiO2 interfaces

Here, we report on the chemistry and structure of 4H-SiC/SiO2 interfaces passivated either by nitric oxide annealing or Ba deposition. Using aberration corrected scanning transmission electron microscopy and spectroscopy, we find that Ba and N remain localized at SiC/SiO2 interface after processing. Further, we find that the passivating species can introduce significant changes to the near-interface atomic structure of SiC. Specifically, we quantify significant strain for nitric oxide annealed sample where Si dangling bonds are capped by N. In contrast, strain is not observed at the interface of the Ba treated samples. Finally, we place these results in the context of field effect mobility.

Influence of experimental conditions on atom column visibility in energy dispersive X-ray spectroscopy.

Here we report the influence of key experimental parameters on atomically resolved energy dispersive X-ray spectroscopy (EDX). In particular, we examine the role of the probe forming convergence semi-angle, sample thickness, lattice spacing, and dwell/collection time. We show that an optimum specimen-dependent probe forming convergence angle exists to maximize the signal-to-noise ratio of the atomically resolved signal in EDX mapping. Furthermore, we highlight that it can be important to select an appropriate dwell time to efficiently process the X-ray signal. These practical considerations provide insight for experimental parameters in atomic resolution energy dispersive X-ray analysis.

In-situ real-space imaging of single crystal surface reconstructions via electron microscopy

Here, we report a high temperature in-situ atomic resolution scanning transmission electron microscopy (STEM) study of single crystal surface structure dynamics. With the approach, we gain direct insight into a double layer reconstruction that occurs on the polar SrTiO3 (110) surface. We find that structural details of this reconstruction can be directly attributed to charge redistribution and the thermal mismatch between the surface and the bulk material. Periodic surface defects, similar to dislocations, are found, which act to relieve stress as the temperature is lowered. Combining STEM observations, electron energy loss spectroscopy, and density functional theory, we highlight the combined role of lattice misfit and charge compensation to determine the structure and chemistry of the observed polar surface reconstruction.

A Deep Convolutional Neural Network to Analyze Position Averaged Convergent Beam Electron Diffraction Patterns

We establish a series of deep convolutional neural networks to automatically analyze position averaged convergent beam electron diffraction patterns. The networks first calibrate the zero-order disk size, center position, and rotation without the need for pretreating the data. With the aligned data, additional networks then measure the sample thickness and tilt. The performance of the network is explored as a function of a variety of variables including thickness, tilt, and dose. A methodology to explore the response of the neural network to various pattern features is also presented. Processing patterns at a rate of u202f∼u202f0.1u202fs/pattern, the network is shown to be orders of magnitude faster than a brute force method while maintaining accuracy. The approach is thus suitable for automatically processing big, 4D STEM data. We also discuss the generality of the method to other materials/orientations as well as a hybrid approach that combines the features of the neural network with least squares fitting for even more robust analysis. The source code is available at https://github.com/subangstrom/DeepDiffraction.

A Convolutional Neural Network Approach to Thickness Determination using Position Averaged Convergent Beam Electron Diffraction

Position averaged convergent beam electron diffraction (PACBED) can accurately measure local specimen thickness with nanometer resolution, which is critical in quantitative scanning transmission electron microscopy (STEM) [1]. These thickness measurements are conducted by pattern matching experiment to simulations, either by least squares fitting (LSF) or by eye. This process, however, can be slow and inaccurate, since only slight contrast feature changes are used be distinguish patterns that are within a few nm. In recent years, convolutional neural networks (CNN) have shown excellent performance in tasks such as self-driving cars, face detection, and text recognition. With multiple convolutional layers in deep, CNN can automatically extract the local features from the images without any feature engineering procedures. The above advantage suggests the promising CNN application in determining thickness from PACBED with different image contrast and patterns.

Absolute-Scale Comparison with Simulation for Quantitative Energy-Dispersive X-Ray Spectroscopy in Atomic-Resolution Scanning Transmission Electron Microscopy

1. School of Physics and Astronomy, Monash University, Melbourne, Australia. 2. School of Applied and Engineering Physics, Cornell University, Ithaca, USA. 3. Monash Centre for Electron Microscopy, Monash University, Melbourne, Australia. 4. Department of Materials Science and Engineering, Monash University, Melbourne, Australia. 5. Department of Materials Science and Engineering, North Carolina State University, Raleigh, USA. 6. School of Physics, University of Melbourne, Melbourne, Australia.

In-situ-by-Ex-situ: FIB-less Preparation of Bulk Samples on Heating Membranes for Atomic Resolution STEM Imaging

Recent advances in in-situ electron microscopy have enabled materials characterization to a new level. Namely, advanced in-situ electron microscopy sample holders integrated with heating, liquid, or gas environmental cells have permitted the investigation of kinetic processes while maintaining atomic resolution [1]. Even with these impressive advancements, however, much in-situ work has been performed on nanoparticles and their supporters by dispersing liquid drops onto grids. For these samples, it is often difficult to precisely align along a low order zone axis, which is required for atomic resolution STEM imaging. For bulk samples prepared by FIB, the Ga contamination and/or the protective layer can cause unwanted reactions that prevent observation of pristine surfaces and interfaces. As a result, few works have successfully shown atomic resolution STEM imaging at temperatures higher than 700 °C for membrane based devices [1, 2]. Given these complications, quantitative, or even qualitative, structural and chemical analysis of bulk samples at high temperature remains challenging.

Numerical modeling of specimen geometry for quantitative energy dispersive X-ray spectroscopy

Transmission electron microscopy specimens typically exhibit local distortion at thin foil edges, which can influence the absorption of X-rays for quantitative energy dispersive X-ray spectroscopy (EDS). Here, we report a numerical, three-dimensional approach to model the geometry of general specimens and its influence on quantification when using single and multiple detector configurations. As a function of specimen tilt, we show that the model correctly predicts the asymmetric nature of X-ray counts and ratios. When using a single detector, we show that complex specimen geometries can introduce significant uncertainty in EDS quantification. Further, we show that this uncertainty can be largely negated by collection with multiple detectors placed symmetrically about the sample such as the FEI Super-X configuration. Based on guidance provided by the model, we propose methods to reduce quantification error introduced by the sample shape. The source code is available at https://github.com/subangstrom/superAngle.