Scanning Transmission Electron Microscopy

Advanced Characterization Methods for Materials Science Applications

Alina Bruma

• Publisher: CRC Press
• Year: 2020
• Print ISBN: 9780367197360
• eBook ISBN: 9780429516160
• Edition: 1
• Topic: Computer Science
• Subtopic: Data Processing
• Index: Computer Science

Scanning Transmission Electron Microscopy is focused on discussing the latest approaches in the recording of high-fidelity quantitative annular dark-field (ADF) data. It showcases the application of machine learning in electron microscopy and the latest advancements in image processing and data interpretation for materials notoriously difficult to analyze using scanning transmission electron microscopy (STEM). It also highlights strategies to record and interpret large electron diffraction datasets for the analysis of nanostructures.

This book:

• Discusses existing approaches for experimental design in the recording of high-fidelity quantitative ADF data
• Presents the most common types of scintillator-photomultiplier ADF detectors, along with their strengths and weaknesses. Proposes strategies to minimize the introduction of errors from these detectors and avenues for dealing with residual errors
• Discusses the practice of reliable multiframe imaging, along with the benefits and new experimental opportunities it presents in electron dose or dose-rate management
• Focuses on supervised and unsupervised machine learning for electron microscopy
• Discusses open data formats, community-driven software, and data repositories
• Proposes methods to process information at both global and local scales, and discusses avenues to improve the storage, transfer, analysis, and interpretation of multidimensional datasets
• Provides the spectrum of possibilities to study materials at the resolution limit by means of new developments in instrumentation
• Recommends methods for quantitative structural characterization of sensitive nanomaterials using electron diffraction techniques and describes strategies to collect electron diffraction patterns for such materials

This book helps academics, researchers, and industry professionals in materials science, chemistry, physics, and related fields to understand and apply computer-science–derived analysis methods to solve problems regarding data analysis and interpretation of materials properties.

1Practical Aspects of Quantitative and High-Fidelity STEM Data Recording

Lewys Jones¹,²

¹Advanced Microscopy Laboratory, Centre for Research on Adaptive Nanostructures and Nanodevices (CRANN), Trinity College Dublin

²School of Physics, Trinity College Dublin

CONTENTS

• 1.1Introduction
• 1.2The Annular Dark-Field Detector
• 1.3Example Applications of Quantitative ADF
• 1.4Simulation Reference, Statistical Decomposition, or Hybrid Analysis Approaches
• 1.5Practicalities of Recording ADF Detector Scans and Choosing ­Camera-Length
  • 1.5.1Confocal or Swung-beam Detector Scanning Approaches
  • 1.5.2Efficiency Response Across the ADF Detector Surface
  • 1.5.3Detector Response Linearity
  • 1.5.4Dropped Gain Versus Dropped Current
  • 1.5.5Choosing Camera Length to Optimize ADF Inner- and Outer-Angle
• 1.6Post-Specimen Flux Distributions and the Effect on Normalization Approach
• 1.7Factors Affecting the Accuracy of Experimental Intensities and Peak Positions
  • 1.7.1Sample Tilt
  • 1.7.2Atomic-column “Cross Talk”
  • 1.7.3Strain Contrast
  • 1.7.4Amorphous Layers and Carbon Background
  • 1.7.5Electron Dose and Pixel Size
  • 1.7.6Cold Field-emission Current Fluctuation and Emission Decay
  • 1.7.7Summary of Sources of Error
• 1.8Environmental Noise and Scanning-Distortion in the Stem
• 1.9Further Multi-Frame Applications of STEM Imaging, Spectroscopy, And 4D-STEM
  • 1.9.1Increasing Image SNR
  • 1.9.2Increasing Spectroscopic SNR
  • 1.9.3Increasing Pixel Density (Digital Super-resolution)
• 1.10Future Possibilities In Quantitative ADF and Multiframe Imaging
  • 1.10.1New Geometries of Dark-field Detection
  • 1.10.2Single Electron Counting Binary ADF Imaging
  • 1.10.3Advances in Scan Patterning and Custom Scan-Design
• 1.11Conclusion
• Acknowledgments
• References

1.1Introduction

Aberration correction in the scanning transmission electron microscope (STEM) allows for the atomic resolution imaging of samples, while the incoherent nature of annular dark-field (ADF) allows for direct interpretation with a Zⁿ contrast relationship (Krivanek et al. 2010). More than only taking “pretty pictures”, since the 1970s (Retsky 1974; Isaacson et al. 1976) there have been significant efforts to quantify this image intensity. At first, this growing field of “quantitative ADF” mostly referred to just the image intensity information; however, as it is often atomic-resolution data that are studied, spatial information such as peak-position shifts and strain mapping have recently started to be studied more. Spatial precision (and the effects of scan-distortion), while not the focus of this chapter, can affect image intensities, so it will be briefly discussed in the context of error analysis.

This chapter is structured as follows. first the hardware of the ADF detector is introduced along with some examples of how quantitative intensity imaging can be used. The mathematics of image normalization is given in the highlighted literature, so this review will instead concentrate on areas of best practice, experiment design, and error analysis in quantitative ADF. Next, the alternative approaches for data analysis are presented, namely either reference simulations, purely statistical approaches, or more recently a robust hybridization of the two. In this discussion of best-practice, Section 1.7 devotes significant discussion to the potential sources of error for practical microscopists. Finally, with quantitative imaging increasingly moving toward picometre-scale spatial precision measurements, Sections 1.8 and 1.9 introduce the use and application of multi-frame recording for the observation, diagnosis, and compensation of environmental noise and scan-distortion.

1.2The Annular Dark-Field Detector

The design of most ADF detectors is fairly similar. To facilitate retraction when not in use, these are usually mounted horizontally on a vacuum bellows. A scintillator inclined at 45° is attached to a light guide so that the signal reaches a horizontally mounted photomultiplier tube (PMT), Figure 1.1 (Kirkland and Thomas 1996). A hole through the scintillator and light guide (often lined with a metal tube to prevent charging) allows the un-scattered bright-field electrons to pass to another detector or spectrometer.

Figure 1.1Left: Schematic of a typical ADF detector showing the hole in the scintillator to allow unscattered electrons through to other detectors or spectrometers. Right: Photograph of a Fischione 3000 ADF detector on a retractable mounting showing the metal-lined tube for unscattered electrons (image credit: (Jones 2016)).

Scattered electrons hit the scintillator and produce optical photons which are detected and amplified at the PMT to produce the final signal. The output voltage from the photomultiplier has a dark-level, set by the amplifier brightness (or offset), which is added to the experimental signal; this combined signal is then amplified further by an amount depending on the amplifier's gain (or contrast) setting. Finally, this PMT voltage is averaged over the dwell-time specified by the user and passed through a so-called analogue-to-digital converter (ADC) (Grillo 2011). The output units are often displayed in “arbitrary counts”; importantly this should not be confused with real electron-counts. We will see later that while modern detector and amplifier systems are sufficiently sensitive to register single electron scattering impacts (Ishikawa et al. 2014), owing to practical limitations this is not presently implemented.

1.3Example Applications of Quantitative ADF

Extracting reliable information from quantitative ADF relies on being able to fix as many experimental factors as possible, ideally to leave only one free parameter to explore. Where thickness is constant (or presumed from extrapolation) quantitative ADF intensity can be used, with extensive simulation, to reveal local compositional variations, Figure 1.2 (Rosenauer et al. 2011). It should be noted that in this case the limiting error will be some combination of the reliability of thickness extrapolation and the effect of de-channeling or static-atomic-displacements.

Figure 1.2Part of an atomic-resolution experimental ADF image of InGaN (a) expressed in contrast units of “fractional beam”. After peak-finding, integration polygons are defined (b), which are used to locally average the image intensity (c). Adjusting for thickness, and through comparison with extensive simulation, the resultant In-concentration map at atomic resolution is produced (d) (image credit: (Rosenauer et al. 2011)).

However, as ADF imaging has a higher intrinsic scattering cross-section than electron energy-loss spectroscopy (EELS) or energy-dispersive x-ray spectroscopy (EDX) (around 100x and 10,000x more respectively), the signal-to-noise ratio (SNR) of quantitative ADF might yield a lower compositional error than these atomic resolution chemical mapping approaches directly, especially for samples that might not tolerate the beam dose of spectroscopy. Where more than one imaging detector is used simultaneously, e.g. low-angle ADF (LAADF) with high-angle ADF (HAADF), and using calibration against other techniques such as x-ray diffraction (XRD), it is possible...

Table of contents

• Cover
• Half Title
• Title
• Copyright
• Dedication
• Contents
• Preface
• About the Editor
• Contributors
• Chapter 1 Practical Aspects of Quantitative and High-Fidelity STEM Data Recording
• Chapter 2 Machine Learning for Electron Microscopy
• Chapter 3 Application of Advanced Aberration-Corrected Transmission Electron Microscopy to Material Science: Methods to Predict New Structures and Their Properties
• Chapter 4 Large Dataset Electron Diffraction Patterns for the Structural Analysis of Metallic Nanostructures
• Index