Quantitative analysis of crystal lattice and defects in nanoscale functional materials by electron diffraction

Crystalline defects are critical to the properties of the material in both desired and undesired ways. In nanoscale functional materials, a small number of defects can change the material performance significantly. Transmission electron microscopy (TEM) has been one of best techniques to study crystalline defects due to its unparalleled spatial resolution. With the rapid advancements in electron detectors, data mining algorithms, and computation power for big data, a new experimental technique in TEM, called scanning electron nanodiffraction (SEND) or four-dimensional (4D) scanning transmission electron microscopy (STEM) or 4D-STEM, is emerging as a powerful way to provide information in both real space and reciprocal space at the same time based on electron nanodiffraction. This thesis aims to develop novel data analysis approaches of SEND datasets for quantitative analysis of crystal lattice and defects, taking advantage of the geometry and intensity of Bragg diffraction, and diffuse scattering in nanobeam diffraction. First, we develop a powerful and versatile technique for lattice strain mapping using SEND. The measurement of strain is based on determining the Bragg peak positions recorded in the diffraction patterns from a local crystal volume. To address the issue of peak broadening from a focused electron probe, we propose a new method based on circular Hough transform to locate the position of non-uniform diffraction disks for high accuracy. Methods for fitting a 2D lattice to the detected disks for strain calculation are described, including error analysis. We demonstrate our technique on a FinFET device for strain mapping at the spatial resolution of 1 nm and strain precision of ~0.03%. By testing on the experimental and simulated four-dimensional diffraction datasets (4D-DDs), the experimental parameters involved in data acquisition and analysis are thoroughly investigated to construct an optimum strain mapping strategy using SEND. Next, techniques for training artificial neural networks (ANNs) and convolutional neural networks (CNNs) using simulated dynamical electron diffraction patterns are described. Using the simulated diffraction patterns as input and trained ANNs, we aim for precise determination of crystal structural properties, such as crystal orientation and local strain. Further, by applying the trained ANNs to 4D-DDs collected using SEND or 4D-STEM techniques, the crystal structural properties can be mapped at high spatial resolution. We demonstrate the ANN-enabled possibilities for the analysis of crystal orientation and strain at high precision and benchmark the performance of ANNs and CNNs by comparing with previous methods. A factor of thirty improvement in angular resolution at 0.01˚ (0.16 mrad) or better for orientation mapping, sensitivity at 0.04% or less for strain mapping, and improvements in computational performance are demonstrated. Lastly, we focus on imaging and characterization of different types of defects. This is demonstrated using SiGe. We explore the possibility to characterize local lattice distortion based on electron diffuse scattering in coherent SEND. Cepstral STEM imaging is proposed and tested on a dislocation core in SiGe to visualize different types of distortion. Using the results from Cepstral STEM, a deep learning-based method is designed to differentiate different types of defects by detecting features in diffuse scattering automatically.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Materials Science & Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2021

Author dc:creator

• Yuan, Renliang

Contributors dc:contributor

• Zuo, Jian-Min
• Eckstein, James N
• Shoemaker, Daniel P
• Huang, Pinshane Y

• Defects
• Electron microscopy
• Electron nanodiffraction
• 4D-STEM
• Scanning electron nanodiffraction
• Orientation mapping
• Strain analysis
• Cepstral STEM
• Machine learning
• Deep learning
• Artificial neural networks
• Semiconductor devices
• Dislocation core
• Lattice distortion

Statement dc:rights

• © 2021 Renliang Yuan

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/110527

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/110527