High-resolution analytical instruments—such as in situ transmission electron microscopes (TEM), scanning transmission electron microscopes (STEM) and electron energy loss spectroscopy (EELS)—are powerful tools that enable insight into the nature of various types of materials. Using particle beams that capture an image, they allow materials to be measured under high temperatures, whether solid, liquid or gas. They also function under other complex electrochemical, optical, and mechanical settings. Yet these instruments have several drawbacks. Specimens can be damaged the longer they sit under the particle beams, and some materials deform too quickly for accurate measurement using current instruments. Further, images generated for analysis require massive amounts of computational power and storage space. And current instruments are challenged to capture changes in microstructure over time, an analysis key to many scientific breakthroughs. Key to all these problems is the number of measurements an instrument must take to accurately render an image of a specimen for analysis. The more an instrument must "sense" about the specimen, the longer it takes, the more damage may be incurred, and the greater the size of the image.

Compressive Sensing is a new mathematical approach that allows samples to be sparsely measured at first, with details accurately recovered later using software. New approaches for compressive sensing acquisition in TEM were developed by researchers at Pacific Northwest National Laboratory. The newly developed Compressing Sensing approach can shorten measurement time, lower file sizes, and track changes over time. For example, with a STEM, Compressive Sensing can reduce the number of measured pixels in an image, speeding image capture and lowering dose from the particle beam. For instruments with video capture, Compressive Sensing increases the effective frame rate of the camera by adding a mask between the specimen and camera. The mask moves at a fixed rate so that a sequence of coded images is integrated into a single frame. Compressive Sensing then arranges these images into a coherent frame. The innovation can also reconstruct full-resolution images from data streams that are incomplete, corrupted, or under-sampled.

With this approach, the number of measurements can be reduced (or a specimen can be measured with lower-resolution hardware), and images still can be reconstructed with high-fidelity. In the case of electron microscopy, this approach allows for complete reconstruction with less than 10 percent of the standard measurements. Dose from the particle beam and image acquisition time can be decreased by at least a favor of five relative to conventional methods. And the approach can be applied without changing microscope operating parameters.

Applicability

Compressive Sensing combines the sensing (measurement) and compression in one operation to increase resolution or speed of any instrument. The approach works with images, video diffraction patterns, signals to create spectra, or other data. It can be applied while the data are being captured or after capture and storage. Researchers could use Compressive Sensing for a variety of ex-situ and in-situ TEM studies, such as following nucleation and growth processes or tracking structural changes during oxidation-reduction reactions. PNNL researchers have implemented the approach for STEM on a specialized research instrument currently being used for client work, while PNNL researchers have implemented video compressive sensing TEM imaging on an FEI Titan 80-300, an advanced analytical field emission STEM.