Towards Low-dose and Fast 4-D Scanning Transmission Electron Microscopy: New Sampling and Reconstruction Approaches

Four-dimensional Scanning Transmission Electron Microscopy (4-D STEM) relies on recording twodimensional (2-D) diffraction patterns for every position of a convergent electron beam In combination with algorithms that solve the phase problem, a 4-D STEM dataset allows the reconstruction of a high-resolution object phase, by exploiting the high amount of redundancy in the data. However, the practical functionality of 4-D STEM is limited in many applications by the relatively slow scan speed imposed by the readout time of the detector (~1ms), longer than the typical dwell time (~10s) in conventional STEM. One approach to circumvent this limitation is to bin the detector, at the cost of degrading the resolvable resolution in the reconstructed object phase. Another alternative is to increase the readout time of the detector and collect only binary counts A new strategy that followed from the successful application of the theory of Compressive Sensing (CS) to STEM [3], showed that a complete set of 4-D STEM measurements can be recovered from simulated incomplete data as if they were recorded by subsampling electron probe positions and the diffraction patterns A phase retrieval algorithm on the recovered 4-D STEM data can then performed to obtain the object phase.