Background

As one of the most powerful atomic-scale imaging methods, high-resolution (HR) transition electron microscopy (TEM) was employed early for mineral analysis and rapidly extended to understand mineral reactions and complex defects (Veblen et al. 1993). In recent years, advances in electron optics, detectors, sample environments as well as data acquisition and analysis methods have enabled the extraction of more information and the study of challenging materials. In particular, the use of cryogenic sample preparation and imaging methods, as well of the development of low-dose imaging techniques, provided opportunities for structural analyses of fully hydrated fine-grained minerals, particularly clays. Our group is collaborating closely with scientists at the National Center for Electron Microscopy in the Molecular Foundry to develop and apply these methods, as well as new directions, to the systems of study in this proposal.

Cryogenic Transmission Electron Microscopy and Tomography

CryoTEM is routinely used to elucidate molecular structures in the life sciences and was the basis of the 2017 Nobel Prize in chemistry. Very fast cryogenic freezing preserves hydrated samples in a near-native state and reduces irradiation damage by decreasing the diffusion length of reactive species generated by inelastic electron scattering (Hattne et al. 2018) making otherwise sensitive samples amenable to real space atomic-scale characterization. The prevailing paradigm in the life sciences is to average hundreds to thousands of images of nominally identical individual molecules as a means to achieve high resolution while circumventing the deleterious effects of electron-induced beam damage on any individual molecule. This paradigm is not applicable for systems in which the objects of interest are unique, which includes most non-biological samples, and is especially true for clay minerals. We have developed an imaging paradigm for nanominerals that takes advantage of the increased dose robustness and native-state preservation of cryo-frozen samples, while enabling the structural resolution to better than 2 Å and tomographic reconstruction in 3D.

Nanomineral structures, even those of layered minerals like clays, are inherently three-dimensional. We have classified defect structures from individual clay particles using cryo electron tomography (CET) to show that there are a hierarchy of pore geometries that are not currently accounted for in conceptual models of clay rich media (Whittaker, Banfield et al. 2020). In particular, topological defects such as dislocations, in which a layer terminates within a particle forcing adjacent layers to bend around it, give rise to pores with diameters that are larger than the equilibrium basal spacing in neighboring interlayers, but smaller than what are conventionally considered to be ‘macropores’, i.e., spaces in between particles. The diffusivities of charged species are highly dependent on the pore diameter, which suggests that these pore structures may play a heretofore unanticipated role in solute transport in clay-rich media.

Capability Development

Electron Multiple Scattering Correction for Tomography Reconstruction

Tomographic methods for the reconstruction of three dimensional objects like cells from two dimensional images are well-established in the life sciences. However, the high amplitude contrast that makes cryoTEM a powerful technique for visualizing hydrated cations in the interlayer of clay particles also limits the applicability of traditional tomographic reconstruction techniques that do not explicitly treat multiple scattering that is common in ‘thick and heavy’ environmental samples. In collaboration with Dr. Colin Ophus at the National Center for Electron Microscopy (NCEM) and Dr. Laura Waller's computational imaging group at UC-Berkeley, we are applying a multiple-scattering tomographic reconstruction algorithm (Ren et al. 2020) to reveal the 3D structure of hydrated clay particles at dimensions smaller than the thickness of an individual layer. With improvements in data processing undertaken in the upcoming proposal period, we expect to achieve lattice resolution (<5 Å) of each of the tens to hundreds of clay mineral layers within a volume of approximately 1 µm x 700 nm x 250 nm. Geometry-dependent microscopic theories of clay layer interactions will then be applied directly to authentic clay mineral structures to determine intermolecular forces that underlie these structures. This will allow for the explicit determination of properties such as permeability, swelling pressure, and shear modulus for which the microscopic origins are currently simulated or inferred from ensemble measurements. This imaging paradigm holds great promise for the atomic scale characterization of particles and interfaces of broad geological significance, including defects, cracks, interfaces, and hierarchical structures.

Information Extraction from Large Tomographic Datasets

The goal of the tomographic imaging project over the three year proposal period is to accurately predict macroscopic properties from microscopic tomographic models. In principle, tomographic data sets contain sufficient multiscale information to accomplish this goal. However, the wealth of information contained within a CET dataset, which is typically tens to hundreds of gigabytes, exceeds current capabilities to analyze it in full. Two new directions arise for expanding the scope of CET data analysis. Initial steps will focus on the automated segmentation of clay mineral layers from CET using deep learning algorithms implemented in Weka (open source) and Dragonfly™ to create multi-scale structural models that can integrate with complimentary characterization techniques such as XRD. Structural models of segmented layers will be used generate three-dimensional pair distance distribution functions that will inform new theories for X-ray scattering and diffraction based on observed microstructures. Establishing a direct link between microscopic features and macroscopic measurements will allow for detailed analysis of clay mineral microstructures using readily available techniques such as XRD without requiring CET to be acquired for each condition of interest. In an orthogonal effort, nanoscale information about the location of isomorphic substitutions, structure of the EDL, effect of curvature and strain on EDL structure, and relative orientation of interacting layers will be determined by in-painting atomic-resolution information using learned features from the data. Average structural units will be calculated using sub-tomogram averaging implemented in IMOD (Mastronarde & Held 2017) and RELION 3.0 (Scheres 2012). These average features will be classified into distinguishable structural motifs and refined iteratively to generate a basis set of atomic-resolution clay mineral unit structures.

Four-Dimensional STEM Ptychography

The dose-efficient technique 4D–STEM ptychography could help to beat the trade-off between SNR and beam damage (Fang et al. 2019). In 4D–STEM, a 2D pixelated detector is used to record the diffraction pattern (Fig. 2), rather than an ADF detector integrating parts of the diffraction pattern and giving one signal at each probe position. One problem of 4D-STEM is that the commercialized 2D pixelated detector has a readout time of 1 ms (1 kHz), which significantly limits its application for clay minerals at the atomic scale because of sample drift caused by sample charging and long readout time. Recently Lawrence Berkeley National Laboratory has developed a 4D-camera working at a very fast readout speed (Ciston et al. 2019), 87 kHz. Using this new 4D-camera, we can record the diffraction pattern at each probe position with a dwell time as short as 12 μs. From the recorded 4D dataset, 2D for scanning probe plus 2D for diffraction pattern, the projected potential of the structure could be reconstructed using the ptychography method, which may be the most effective way to observe light elements like oxygen (Yang et al. 2017). Using this cutting-edge STEM technique, we anticipate resolving all the atomic columns along both [010] and especially [010]

Applications to Proposal Objectives

Electron microscopy methods will have numerous applications through all the proposal topics. Key examples include:

Topic 1 High-resolution Tomography of Nanopores in Biogenic Carbonates

Topic 2 Hydrated Structure of Fresh, Aged and Healed Single Cracks

Topic 3 Defect Stacking Structures of 2:1 and 1:1 Clays

References