When 10:00 AM - 12:00 PM Sep 01, 2023
Where 3158 HH Dow (Pod Room)
Add event to calendar vCal

PhD defense: "Breaking Dose Limitations for High-Resolution Spectroscopy with Fused Multi-Modal Electron Microscopy"

Jonathan Schwartz
Hovden group

In the realm of modern high-performance materials, intricate 3D heterogeneous architectures with precise chemistry manipulation at nanometer to atomic scales have become feasible. With the development of aberration-correctors in scanning/transmission electron microscopes (S/TEM), material structures can now be routinely imaged at sub-angstrom resolutions. However. The necessity of high electron doses for capturing inelastic scattering events has hindered high-resolution spectroscopy, impeding the measurement of detailed chemistry. This fundamental constraint particularly affects the adoption of 3D chemical tomography as few specimens can withstand the intense beam irradiation. 

Fortunately, the emergence of modern computational frameworks has paved the way for low-dose specimen imaging. By framing the recovery process as an inverse problem, it becomes possible to measure chemistry at high resolutions with significantly reduced doses. This study addresses the challenge of characterizing material chemistry in 2D and 3D at sub-nanometer spatial resolutions – within the tolerable dose limits of most materials. This advancement broadens the applicability of such analyses to a wider range of materials. 

This study introduces a pioneering framework for multi-modal data fusion of signals within electron microscopes. As a result, we can harness synergistic advantages of each modality by fostering effective communication among correlated multi-modal signals. This innovative paradigm offers a pathway to augment our comprehension of intricate material chemistry with minimal dose. Specifically, chemically sensitive maps with low signal-to-noise ratios (SNR) can benefit significantly from leveraging high-SNR structural signals acquired through elastic scattering events (e.g. high-angle annular dark-field micrographs). Notably, this enhancement yields a 2D improvement in chemical spectroscopy, enabling the recovery of individual elements and a 3D advancement that reveals comprehensive 3D distributions, exposing interfacial chemistry. 

A diagram of a multi-modal data fusion

Description automatically generatedConversely, the time required to visualize a material's 3D structure through any S/TEM signal is often prolonged, largely due to the complexities surrounding dataset processing. Beyond the challenges of data acquisition, processing a single electron tomography dataset may entail hours to days for a solitary visualization. We address these obstacles by constructing a centralized pipeline that handles the computational load, data processing, and visualization at the touch of a button. Coupled with contributions to the widely-used open-source volumetric rendering software at tomviz.org, electron tomography can now unfold concurrently with data collection, empowering researchers to discern 3D structures in real-time while operating the microscope.

The advancements stemming from this work equip 3D characterization to seamlessly encompass both structural and chemical attributes, thereby enabling researchers to concentrate on resolving the intricate challenges in materials science.