Technique Development

Technique Development

Many of our research efforts focus on developing new methods for real-time studies using electron microscopy methods.  The philosophical motivation for this is described in an article on how science advances by A.M. Schneider entitled “Four stages of a scientific discipline: four types of scientists“. He posits that there are four stages of research:

  1. Identification of the object of research
  2. Development of tools/instruments to investigate the underlying phenomena
  3. Investigation of the research by the developed tools, and
  4. The codification and refinement of knowledge.

Below, you will find short descriptions of some of our work concerning Stage 2.  Stage 3 research work is described on the Research page.

Echem-Liquid cell

Liquid-cell photo/electrochemistry

We are utilizing the technique of liquid-cell electron microscopy to understand the processes of electrocatalysis and photo-electrochemsitry. With this approach it is possible to image materials in liquid environments at the nanometer to atomic level during reaction. The technique of liquid cell microscopy was pioneered by our long time collaborator Frances Ross (MIT). 

Our recent technique developments include:

  • Development of fully quantitative electrochemical experimentation, with colleagues at Hummingbird Scientific.
  • Methods that allow routine, high resolution analytical microscopy during electrochemical experimentation.
  • Development of methods for photoelectrochemistry, again with colleagues at Hummingbird Scientific.

 

Atmospheric pressure catalysis

In atmospheric pressure microscopy, we use closed-cell methods similar to those for liquid microscopy to provide both high temperatures and atmospheric pressure. This allows us to understand how the size, structure and composition of supported heterogeneous catalyst change in reactive environments.

Machine Learning

We apply state-of-the-art machine learning techniques to problems in electron microscopy. Initial work focuses on developing deep-learning methods for binary image segmentation to aid the processing of electron micrographs. While our modern direct electron detection cameras enable imaging hundreds of frames per second, the amount of data we extract is limited by the number images that can actually be processed – our recent results have shown that applying a  convolutional neural network to images of supported nanoparticles can reduce the time required to identify and measure particles from over 12 hours per image to merely 0.5 seconds per image.