Final Report Summary - 3DIMAGE (3D Imaging Across Lengthscales: From Atoms to Grains)
To understand the complex interplay between structure and composition of modern materials and devices and their underlying function and properties requires characterisation techniques that can explore samples across many lengthscales. Electron tomography (3D imaging) has traditionally allowed the morphology of materials to be studied with nanometre resolution. This project has built upon this by expanding tomographic methods to be applied from atomic through nanoscale to mesoscale resolution and, through combining with a number of different analytical techniques, enabled the determination of 3D chemistry, electronic and optical properties and crystallography down to the nanoscale. The research was split into three workpackages (WPs) that focus broadly on three lengthscales and a fourth WP focussed on algorithmic development, image and spectral processing. These boundaries were blurred in practice and all workpackages were interlinked. Below is a brief summary of the work undertaken and some of the more important results and developments:
In work package (WP) A, ‘atomic scale tomography’, we used STEM HAADF imaging to investigate nanoscale catalysts, in particular GaPd2, a candidate catalyst with an isolated single-site for hydrogenation reactions. We studied decahedral and icosahedral particles and were able to achieve new insights into how ‘5-fold’ twinning can be better accommodated in orthorhombic crystals. Direct imaging of small atomic clusters and of single atoms has also been achieved using STEM HAADF imaging in movie mode. We developed novel video de-noising methods to enable accurate tracking of single atoms, dimers and trimers on a variety of substrates. Metal atom diffusion on both silicon and graphene oxide surfaces was studied. On disordered graphene oxide, surface atoms diffused in an anomalous fashion (sub-diffusive and non-ergodic) whereas on a Si (110) surface normal (Brownian-like) motion was observed. In addition, we studied the question of reconstruction fidelity in atomic-resolution tomography, looking at both crystalline and amorphous material, through an entropy (information)–based approach.
In WPB, ‘nanoscale tomography’, we made progress towards true 3D nano-metrology developing more automated unbiased tomogram segmentation. As an example, the size, shape and distribution of a catalyst agglomerate with >2000 nanoparticles was determined, a task that would be nearly impossible to do manually. We developed analytical electron tomography combining EELS and EDX with 3D imaging (tomography) to produce 3D nanoscale maps of composition and chemistry. Moreover we applied EELS tomography to understand the 3D plasmonic behaviour of metal nanoparticles, showing for the first time how these techniques can be used to determine the underlying geometric eigenmodes and link the near-field response determined by EELS with the far-field optical spectra. Other nanoscale structures were studied also by EELS and tomography to investigate the Starck effect in III-V nanowires, metal-organic frameworks, and trimers and tetramers. We have also developed a new crystallographic form of tomography to enable a full 3D reconstruction of the local lattice orientation and dimension, in principle, at every real space voxel. By using machine learning methods such enormous data sets become tractable and important 3D crystallographic information recovered at the nanoscale including orientation relationships and local 3D strain.
In WPC, ‘mesoscale tomography’, we combined dual beam SEM-FIB, to reveal 3D morphology, with EDX to achieve 3D quantitative chemical mapping. New software was developed to enable both low energy and high energy x-rays to be used to correct for x-ray absorption and to improve quantification. This was applied to a study of Ni base superalloys, of possible use in next-generation turbines, where subtle compositional fluctuations were revealed in gamma prime precipitates. Similar methods were also used to understand the local structure and composition in certain magnetic minerals: we were able to reconstruct the 3D microstructure with sufficient accuracy to allow magnetic simulations of that structure and ultimately to help understand the origin of the earth’s magnetic field.
In WP D we have been successful in developing new reconstruction methods for 3D imaging, focussing in particular on compressed sensing techniques. These act as mathematical frameworks onto which we can add important constraints (prior information) about the sample or the acquisition that enables the final reconstruction to be of much higher fidelity than would be the case with conventional techniques. Moreover, we can reduce the dose/time by an order of magnitude and recover a 3D tomogram of comparable quality as one with full sampling. We have also implemented new machine learning techniques that have proven to be highly successful and developed a suite of software that is now used world-wide. We have developed new software for video de-noising and new analytical tools that allow atomic cluster and single atom dynamics to be studied quantitatively, and to make important conclusions about the nature of the diffusion of atoms on a variety of surfaces.
In work package (WP) A, ‘atomic scale tomography’, we used STEM HAADF imaging to investigate nanoscale catalysts, in particular GaPd2, a candidate catalyst with an isolated single-site for hydrogenation reactions. We studied decahedral and icosahedral particles and were able to achieve new insights into how ‘5-fold’ twinning can be better accommodated in orthorhombic crystals. Direct imaging of small atomic clusters and of single atoms has also been achieved using STEM HAADF imaging in movie mode. We developed novel video de-noising methods to enable accurate tracking of single atoms, dimers and trimers on a variety of substrates. Metal atom diffusion on both silicon and graphene oxide surfaces was studied. On disordered graphene oxide, surface atoms diffused in an anomalous fashion (sub-diffusive and non-ergodic) whereas on a Si (110) surface normal (Brownian-like) motion was observed. In addition, we studied the question of reconstruction fidelity in atomic-resolution tomography, looking at both crystalline and amorphous material, through an entropy (information)–based approach.
In WPB, ‘nanoscale tomography’, we made progress towards true 3D nano-metrology developing more automated unbiased tomogram segmentation. As an example, the size, shape and distribution of a catalyst agglomerate with >2000 nanoparticles was determined, a task that would be nearly impossible to do manually. We developed analytical electron tomography combining EELS and EDX with 3D imaging (tomography) to produce 3D nanoscale maps of composition and chemistry. Moreover we applied EELS tomography to understand the 3D plasmonic behaviour of metal nanoparticles, showing for the first time how these techniques can be used to determine the underlying geometric eigenmodes and link the near-field response determined by EELS with the far-field optical spectra. Other nanoscale structures were studied also by EELS and tomography to investigate the Starck effect in III-V nanowires, metal-organic frameworks, and trimers and tetramers. We have also developed a new crystallographic form of tomography to enable a full 3D reconstruction of the local lattice orientation and dimension, in principle, at every real space voxel. By using machine learning methods such enormous data sets become tractable and important 3D crystallographic information recovered at the nanoscale including orientation relationships and local 3D strain.
In WPC, ‘mesoscale tomography’, we combined dual beam SEM-FIB, to reveal 3D morphology, with EDX to achieve 3D quantitative chemical mapping. New software was developed to enable both low energy and high energy x-rays to be used to correct for x-ray absorption and to improve quantification. This was applied to a study of Ni base superalloys, of possible use in next-generation turbines, where subtle compositional fluctuations were revealed in gamma prime precipitates. Similar methods were also used to understand the local structure and composition in certain magnetic minerals: we were able to reconstruct the 3D microstructure with sufficient accuracy to allow magnetic simulations of that structure and ultimately to help understand the origin of the earth’s magnetic field.
In WP D we have been successful in developing new reconstruction methods for 3D imaging, focussing in particular on compressed sensing techniques. These act as mathematical frameworks onto which we can add important constraints (prior information) about the sample or the acquisition that enables the final reconstruction to be of much higher fidelity than would be the case with conventional techniques. Moreover, we can reduce the dose/time by an order of magnitude and recover a 3D tomogram of comparable quality as one with full sampling. We have also implemented new machine learning techniques that have proven to be highly successful and developed a suite of software that is now used world-wide. We have developed new software for video de-noising and new analytical tools that allow atomic cluster and single atom dynamics to be studied quantitatively, and to make important conclusions about the nature of the diffusion of atoms on a variety of surfaces.