High Speed/Low Dose Analytical Electron Microscopy with Dynamic Sampling

A new and extraordinarily advanced microscopy technique discovery by researchers at NUANCE speeds up the way materials can be imaged at very small length scales. The breakthrough benefits microscopists and investigators who heavily rely on imaging.

The software package developed by researchers greatly reduces the required amount of sample exposure to an electron beam while maintaining great accuracy.

Karl Hujsak, main author of the paper High Speed/Low Dose Analytical Electron Microscopy with Dynamic Sampling published in Science said “modern microscopes can give us an element by element picture of their structure,  but at the cost of needing hour or days to form a single image. By using machine learning algorithms to make decisions about where to image and how to perform the experiment, we can take the same image an order of magnitude faster than a human operator. This will allow us to study a whole new set of materials that simply weren’t stable for long enough to get a good picture.”

The discovery benefits any microscope that can make multiple single measurements in a row by adding on software. And like the self-driving car, the self-driving microscope will allow researchers to understand  materials with weak bonding that fluctuate faster than a traditional image could be recorded.  This could directly impact our ability to design next-generation catalytic and drug delivery systems at the atomic scale, using algorithms to image them without perturbing their structure

 “Allowing intelligent computers to perform experiments for us enables us to do more as scientists,” Hujsak added.

It is expected that Multi-Objective Autonomous Dynamic Sampling (MOADS) and similar supervised dynamic sampling approaches may open the exploration of large area analytical maps or the imaging of beam reactive materials not previously thought feasible.