Strategized subsampling minimizes dose in angstrom resolution STEM

High resolution imaging in electron (or ion) beam systems often comes at the risk of damaging the sample with the charged particles before a meaningful image can be obtained. The beam’s impact at a given location is typically determined by the number of charged particles hitting that location. Dosage is typically optimized by tuning beam parameters that change the dose for each pixel. Work described in Applied Physics Letters acquires the same resolution at smaller overall doses to the sample by a simultaneous denoising and intentional recovery of missing data in scanning electron transmission microscopy (STEM).

A specific type of compressive sensing, called inpainting, can be used for both low-dose images, where many pixels exhibit zero particle counts, and randomly subsampled images. Both of these methods produce equivalent images in terms of dose, resolution and contrast, but random subsampling can be significantly faster as it skips most pixels. Moreover, the subsampling can be adapted sequentially to minimize dose and maximize resolution. Co-author Andrew Stevens describes this as “intelligent subsampling.” STEM measurements made using this inpainting technique can achieve an order of magnitude better imaging performance than conventional methods, permitting angstrom level resolution at doses below the 10 electron per squared angstrom dose threshold for biological samples.

Building on the assumption that data will be imperfect, undersampling capitalizes on this unavoidable reality with deliberate imperfections (missing pixels) designed to acquire the most information with the fewest pixels. This produces data that requires software processing for image reconstruction, which shifts a portion of the sensing burden from hardware to software.

The authors present comparisons of reconstructed images based on real and simulated data for GaAs, ZnSe, NiTiO₃, and a metal organic framework (MOF) to demonstrate the effectiveness of the subsampling refinement. Its application is generalizable to other dose-sensitive systems such as focused ion beam (FIB) microscopy.

Source: “A sub-sampled approach to extremely low-dose STEM,” by A. Stevens, L. Luzi, H. Yang, L. Kovarik, B. L. Mehdi, A. Liyu, M. E. Gehm, and N. D. Browning, Aplied Physics Letters (2017). The article can be accessed at https://doi.org/10.1063/1.5016192 .