Efficient and quantitative phase imaging in two‐ and three‐dimensions using electron ptychography in STEM

Abstract

Historically, the scanning transmission electron microscope (STEM) has not been widely used for phase contrast imaging because the small bright‐field detector required makes use of only a small fraction of the incident electrons and is therefore inefficient with respect to dose. This limitation has hindered the efficient imaging of light elements in STEM. Alternative modes also have limitations. For example, annular dark‐field (ADF) imaging of graphene only makes use of a few percent of the incident electrons, and annular bright‐field imaging (ABF) requires lens aberrations to form an effective phase plate to get contrast from weakly scattering objects. Electron ptychography in the STEM was first demonstrated more than 20 years ago in the context of improving image resolution [1]. At that time, the image field of view was restricted by the limitations of the camera technology and data handling technology. Here we make use of the pnCCD (S)TEM camera, a direct electron pixelated detector from PNDetector, mounted on the JEOL ARM200‐CF aberration corrected microscope. The detector has a grid of 264x264 pixels and operates at a speed of 1000 frames‐per‐second (fps). The detector can achieve a speed of up to 20,000 fps through binning/windowing. ADF images can be recorded simultaneously, as shown by the schematic in Fig. 1. The resulting 4D data set is formed of a series of coherent convergent beam diffraction patterns recorded as a function of illuminating probe position. Here we explore how the bright‐field and dark‐field regions of scattering can be used to enhance the capabilities of STEM. We compare a range of methods that can be used to form the phase image from this data set, including single side‐band [2,3], Wigner distribution deconvolution [4] (used to produce Fig. 2) and ePIE [5]. Phase imaging using ptychography has a relatively simple transfer function [3] and also provides an inherent filter of image noise without reducing the signal strength to form high quality phase images (Fig. 2). Furthermore, the four‐dimensional data set is highly redundant and it is possible to detect and correct for residual aberrations in the image. The ability to deconvolve lens aberrations can further be used to extract three‐dimensional information from a single STEM image acquisition scan. This is achieved by reconstructing the phase image at a specific depth in the sample, which can be performed even though the microscope may not have been focused at that depth (Fig. 3). Finally, we explore the potential for using information outside the bright‐field disc to enhance STEM imaging.

Convolutional Neural Network for Extracting 3D Point Clouds of Fibrous Web From Multi-Focus Images

“…The optical coherence tomography is the most common way that uses the sequential images of an object captured on various focal planes/depths to reconstruct the 3D surface image [4]- [6]. Quantitative phase imaging (QPI) can developed to deal with transparent and translucent objects in optical microscopy [7], [8]. Based on QPI, LED matrix illumination was utilized to capture images under different beam angles by controlling the LED arrays [9]- [11].…”