High-Dimensional Compressive Scanning Transmission Electron Microscopy
Transmission electron microscopy (TEM) is entering the era of big data, with instruments thought of not so much as image-collection devices as information¬-collection devices. Not so many years ago, spatial resolution was king; TEM technology competed to produce images at ever higher resolution. But this trend has largely played out. Today, with a modern TEM, the instrument’s spatial resolution is rarely the bottleneck that determines whether a particular experiment is possible. Instead, things like electron beam damage, data throughput (including time resolution), informational efficiency, and operational flexibility are what need to be improved. These factors all interact in complex ways that differ with each experiment. But one common theme is a growing recognition that TEM data acquisition needs to be smarter than it has been. One cannot just scale up beam brightness, scan speeds, and detector readout rates indefinitely without running into fundamental problems of space charge, signal to noise ratio, and beam damage. One needs to make better use of the microseconds and electrons that one already has. This brings us to the linked concepts of compressive sensing (CS), adaptive scanning, and sparsity-based inpainting, denoising, and data analysis. CS and its relatives use the reliable presence of patterns in real-world data to create new approaches for acquiring data in more efficient and less brute-force ways. The same mathematical patterns that let video streams be compressed 10-fold or more while losing almost no viewer-relevant information can also be used in the design of clever data acquisition schemes. CS schemes perform data compression in the analog domain, before the signal even reaches the analog-to-digital convertor bottleneck. In the previous phases of the present SBIR project, IDES developed a completely new CS operating mode for electron microscopy and demonstrated it using beta versions of a new product line called Relativity®. The system uses electrostatic subframing (ES) to subdivide a camera, plus a high-speed digital pattern generator to switch among the subframes in an arbitrary preprogrammed sequence synchronized with the camera, scan generator, in situ sample drivers, and so forth. This allows a conventional low-speed TEM camera to acquire kHz-scale video in burst mode, with up to 100 subframes of video compressively encoded into a single camera acquisition. ES thus can bring new life to old microscopes with old camera systems, but it can also (as shown in Phase IIA) be used with a new generation of high-speed cameras to reach even higher performance, including 19,200 subframe-per-second continuous atomic resolution video acquisition with no rolling readout or other timing artifacts. In Phase IIA, outstanding tasks related to data analysis software, system integration, and development of a new electrostatic beam blanker were completed, and post-beta Relativity® systems are now being sold in partnership with JEOL, Ltd. This report describes these tasks, including their state at the end of Phase II and the results produced during Phase IIA. Despite the project ending prematurely due to the acquisition of IDES, Inc., all major tasks were completed successfully and several new product lines have been introduced as a result.
- Research Organization:
- IDES, Inc.
- Sponsoring Organization:
- USDOE Office of Science (SC), Basic Energy Sciences (BES)
- DOE Contract Number:
- SC0013104
- OSTI ID:
- 1614920
- Type / Phase:
- SBIR (Phase I)
- Report Number(s):
- DOE-IDES-13104-9
- Country of Publication:
- United States
- Language:
- English
Similar Records
TEM Video Compressive Sensing
Ultra-High Frame Rate Direct Detector for EBSD and TKD