Bayesian Experimental Design for in situ (Scanning) Transmission Electron Microscopy – BED-TEM

Introduction

The proposed project, Bayesian Experimental Design for in situ (Scanning) Transmission Electron Microscopy (BED-TEM), addresses challenges in conducting in situ (S)TEM experiments, particularly in selecting optimal stimuli parameters.

Background

While recent advancements in microelectromechanical systems (MEMS) have expanded stimulus options, determining appropriate values remains elusive, often relying on time-consuming trial and error.

Project Goals

The project aims to introduce a software platform integrating Bayesian experimental design with (S)TEM image analysis, facilitating efficient parameter selection. This platform comprises:

• User interface
• Image processing
• Experimental design modules

These components offer a streamlined workflow and enhanced usability.

Methodology

Leveraging expertise in machine learning and (S)TEM, the project seeks to bridge the gap between complex experimental setups and the practical application of Bayesian methods.

Innovation

The envisioned breakthrough lies in transforming offline, iterative parameter determination into an online, iterative process, revolutionizing how in situ experiments are conducted and accelerating materials research and development.

Risks and Mitigation Strategies

The project also poses high risks, particularly in:

1. Adapting machine learning to (S)TEM data
2. Ensuring market demand for the proposed software

However, mitigation strategies include:

• Iterative development
• Collaboration with experts
• Continuous user feedback to align the solution with customer needs and commercial viability

Conclusion

Ultimately, BED-TEM promises to reshape materials science by enabling precise characterization of material behaviors at the nanoscale under real-world conditions, with potential applications spanning electronics, aerospace, and beyond.