This image shows live capture and inpainting of sub-sampled 4D STEM data of an interface between silicon and germanium telluride. In this live acquisition, we simultaneously image the LAADF (a-c) and ABF (d-e) signals and inpaint them on-the-fly. We start with a pristine boundary (a, d) acquired at 200s after finding the region and correcting low-order aberrations, then increase the magnification to induce beam damage at the interface (b, e). This causes the interface to breakdown and for tellurium to diffuse across the boundary. The final damaged section can be seen after demagnifying (c, f). Data acquired at CNR-IMM, Catania, Italy. Sample provided by CNR-IMM, Catania, Italy.
The CNR Institute for Microelectronics and Microsystems (CNR-IMM), one of Europe’s leading semiconductor research labs, has started using subsampling software from SenseAI to speed up their electron microscopy imaging, reduce beam damage and control the amount of data used.
Using SenseAI, CNR-IMM has been able to achieve a faster alignment with a lower electron dose, using 10% of the original data. They rapidly test different parameters to get the maximum output from their analysis, crucial for any semiconductor laboratory.
CNR-IMM is an institute of the Italian National Research Council and works with some of the largest semiconductor factories in Europe. CNR-IMM uses STEM and low-energy SEM to research the structural characterisation of materials, support electron crystallography, and carry out customised imaging. They also perform cryogenic TEM measurements on radiation sensitive materials such as graphene or polymers.
SenseAI, a spinout from the University of Liverpool, uses proprietary ‘Compressed Sensing’ algorithms which only samples a fraction of the data without loss of any inherent information. This generates images faster with significantly reduced beam damage and up to 100x less data. Most importantly, there is no loss of integrity of the images.Giuseppe Nicotra, Head of Sub-Ångstrom Electron Microscope LAB at CNR-IMM says: “In the semiconductor industry many specimens are very beam-sensitive. With SenseAI we can work at previously unachievably low doses to preserve sample integrity and benefit from superior data.
“We can perform analysis across SEM, 2D and 4D-STEM in real time. There's nothing worse when you acquire a 4D-STEM data set and then you're unable to know how good it is until you've processed and analysed it, and ten have to re-acquire the datasets. This is a lengthy process consuming huge amounts of data. With SenseAI you can see 4D-STEM images live and make adjustments on the fly.
“Electron microscopy datasets are huge, even generating these large datasets requires very long acquisition times. In our experiments using SenseAI, we are able to subsample using just 10% or less of the original data. Maintaining large data volumes in a data centre is costly but we can now preserve and easily handle all of the required information using a lot less storage.“For electron microscope operators, it's really important to have a very reactive system, especially when you are tuning imaging parameters - you need very fast feedback. Tuning imaging parameters with SenseAI was great, and I’m really impressed at how SenseAI denoises, refines and reconstructs very noisy images.”Visit senseai.vision for more information.