Uncovering material deformations via machine learning combined with four-dimensional scanning transmission electron microscopy

Shi, Chuqiao; Cao, Michael C.; Rehn, Sarah M.; Bae, Sang‐Hoon; Kim, Jeehwan; Jones, Matthew R.; Muller, David A.; Han, Yimo

doi:10.1038/s41524-022-00793-9

Cited by 25 publications

(17 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Electron microscopy 297,298 is an important method by which to study structures and morphology characteristics from the micrometer to the angstrom scale, and is widely used in the characterization of various materials. [299][300][301] In recent years, an advanced topic 302,303 has combined electron microscopy with ML, for which unsupervised learning 304 is particularly prominent.…”

Section: Microstructure Characterization Assistancementioning

confidence: 99%

Methods, progresses, and opportunities of materials informatics

Li,

Zheng

2023

InfoMat

View full text Add to dashboard Cite

As an implementation tool of data intensive scientific research methods, machine learning (ML) can effectively shorten the research and development (R&D) cycle of new materials by half or even more. ML shows great potential in the combination with other scientific research technologies, especially in the processing and classification of large amounts of material data from theoretical calculation and experimental characterization. It is very important to systematically understand the research ideas of material informatics to accelerate the exploration of new materials. Here, we provide a comprehensive introduction to the most commonly used ML modeling methods in material informatics with classic cases. Then, we review the latest progresses of prediction models, which focus on new processing–structure–properties–performance (PSPP) relationships in some popular material systems, such as perovskites, catalysts, alloys, two‐dimensional materials, and polymers. In addition, we summarize the recent pioneering researches in innovation of material research technology, such as inverse design, ML interatomic potentials, and microtopography characterization assistance, as new research directions of material informatics. Finally, we comprehensively provide the most significant challenges and outlooks related to the future innovation and development in the field of material informatics. This review provides a critical and concise appraisal for the applications of material informatics, and a systematic and coherent guidance for material scientists to choose modeling methods based on required materials and technologies.

show abstract

Section: Microstructure Characterization Assistancementioning

confidence: 99%

Methods, progresses, and opportunities of materials informatics

Li,

Zheng

2023

InfoMat

View full text Add to dashboard Cite

show abstract

“…This work reports a structurally agnostic, data-driven workflow for domain mapping, which requires no prior knowledge of possible crystal structures and is equally as effective for identifying amorphous regions with little to no crystalline diffraction character. Techniques based on feature extraction with non-negative matrix factorisation (NMF) 21 and hierarchical k-means clustering to group similar diffraction signal 22 have recently been reported. The workflow, discussed within, leverages a variational autoencoder (VAE) to identify the variance and similarity within the diffraction data.…”

Section: Introductionmentioning

confidence: 99%

Versatile domain mapping of scanning electron nanobeam diffraction datasets utilising variational autoencoders

et al. 2023

View full text Add to dashboard Cite

Characterisation of structure across the nanometre scale is key to bridging the gap between the local atomic environment and macro-scale and can be achieved by means of scanning electron nanobeam diffraction (SEND). As a technique, SEND allows for a broad range of samples, due to being relatively tolerant of specimen thickness with low electron dosage. This, coupled with the capacity for automation of data collection over wide areas, allows for statistically representative probing of the microstructure. This paper outlines a versatile, data-driven approach for producing domain maps, and a statistical approach for assessing their applicability. The workflow utilises a Variational AutoEncoder to identify the sources of variance in the diffraction signal, and this, in combination with clustering techniques, is used to produce domain maps. This approach is agnostic to domain crystallinity, requires no prior knowledge of crystal structure, and does not require simulation of a library of expected diffraction patterns.

show abstract

“…Machine learning (ML) techniques have been widely applied in electron microscopy for applications such as atom localization, 1-3 defect identification, [4][5][6] image denoising, [7][8][9] determining crystal tilts and thickness, [10][11][12] classifying crystal structures, 13,14 optimizing convergence angles, 15 identifying Bragg disks, 16 visualizing material deformations, 17 automated microscope alignment, 18 and many others. Several recent reviews [19][20][21] provide an overview of new and emerging opportunities at the interface of electron microscopy and ML.…”

Section: Introductionmentioning

confidence: 99%

Using CycleGANs to Generate Realistic STEM Images for Machine Learning

Khan¹,

Lee²,

Huang³

et al. 2023

Preprint

View full text Add to dashboard Cite

The rise of automation and machine learning (ML) in electron microscopy has the potential to revolutionize materials research by enabling the autonomous collection and processing of vast amounts of atomic resolution data. However, a major challenge is developing ML models that can reliably and rapidly generalize to large data sets with varying experimental conditions. To overcome this challenge, we develop a cycle generative adversarial network (CycleGAN) that introduces a novel reciprocal space discriminator to augment simulated data with realistic, complex spatial frequency information learned from experimental data. This enables the CycleGAN to generate nearly indistinguishable images from real experimental data, while also providing labels for further ML applications. We demonstrate the effectiveness of this approach by training a fully convolutional network (FCN) to identify single atom defects in a large data set of 4.5 million atoms, which we collected using automated acquisition in an aberration-corrected scanning transmission electron microscope (STEM). Our approach yields highly adaptable FCNs that can adjust to dynamically changing experimental variables, such as lens aberrations, noise, and local contamination, with minimal manual intervention. This represents a significant step towards building fully autonomous approaches for harnessing microscopy big data.

show abstract

Uncovering material deformations via machine learning combined with four-dimensional scanning transmission electron microscopy

Cited by 25 publications

References 35 publications

Methods, progresses, and opportunities of materials informatics

Methods, progresses, and opportunities of materials informatics

Versatile domain mapping of scanning electron nanobeam diffraction datasets utilising variational autoencoders

Using CycleGANs to Generate Realistic STEM Images for Machine Learning

Contact Info

Product

Resources

About