A Practical Reconstruction Method for Three-Dimensional Phase Contrast\n Atomic Electron Tomography

David Ren, Michael Chen, Laura Waller, Colin OphusarXiv 2018 · DOI · arXiv

Abstract: Electron tomography is a technique used in both materials science and structural biology to image features well below the optical resolution limit. In this work, we present a new algorithm for reconstructing the three-dimensional (3D) electrostatic potential of a sample at atomic resolution from phase contrast imaging using high-resolution transmission electron microscopy (HRTEM). Our method accounts for dynamical and strong phase scattering, providing more accurate results with much lower electron doses than current atomic electron tomography (AET) experiments. We test our algorithm using simulated images of a synthetic needle geometry dataset composed of an amorphous silicon dioxide shell around a silicon core. Our results show that, for a wide range of experimental parameters, we can accurately determine both atomic positions and species, and identify vacancies even for light elements and disordered materials.

Introduction

Atomic Electron Tomography (AET) has traditionally relied on Annular Dark Field Scanning Transmission Electron Microscopy (ADF-STEM). While ADF-STEM provides images that are roughly linear projections of the sample's potential, it requires extremely high electron doses that can damage sensitive samples and is susceptible to artifacts from sample motion during scanning.

High-Resolution Transmission Electron Microscopy (HRTEM), which utilizes phase contrast, offers a more dose-efficient alternative. However, phase contrast is highly nonlinear; for samples thicker than a few atomic layers, electrons undergo multiple scattering events (dynamical scattering). This creates complex interference patterns that are difficult to interpret directly. This paper addresses the inverse problem: mathematically "unscrambling" these 2D phase-contrast images to recover the 3D positions and species of every atom in a sample.

Methods

The proposed reconstruction pipeline is built upon three primary pillars designed to handle the complexities of phase-contrast imaging:

  1. Multislice Forward Model: Rather than assuming straight-line propagation, the algorithm treats the sample as a sequence of thin 2D slices. It models the electron wave interacting with each slice and then propagating through free space to the next. This accurately accounts for multiple scattering and strong phase shifts.
  2. Iterative Optimization: The algorithm begins with a blank volume and iteratively updates the estimated potential to minimize the difference between predicted images and actual measurements. The authors found that minimizing the residual of the wave amplitude (the square root of intensity) was particularly effective at handling Poisson noise in low-dose data.
  3. Computational Efficiency: To manage the high computational cost, the researchers implemented "Slice-Binning," combining multiple thin slices into thicker ones during reconstruction. This maintains high 3D resolution by leveraging the redundant information available in tomographic tilt series while significantly reducing calculation time.

Experimental Results

The method was evaluated using a synthetic needle model consisting of a crystalline silicon (Si) core and an amorphous silicon dioxide ($SiO_2$) shell, containing over 150,000 atoms. This model provided a rigorous test for both ordered and disordered regions, as well as species differentiation.

Precision and Species Identification

Under standard conditions, the algorithm achieved a mean 3D position error of only 12–14 picometers (pm). By analyzing the intensity of the reconstructed potential peaks, the method successfully distinguished between silicon and oxygen atoms, even in the disordered amorphous shell where bond lengths were accurately resolved to approximately 1.6 Å.

Dose Efficiency and Vacancy Detection

A critical finding was the algorithm's performance at low electron doses. Even at a dose of 7,000 $e^-/Å^2$—orders of magnitude lower than typical AET experiments—the algorithm correctly identified over 95% of the atoms. Furthermore, the use of Total Variation (TV) regularization allowed for the successful detection of single-atom vacancies in both the crystalline and amorphous regions by suppressing background noise without blurring sharp atomic peaks.

Discussion and Conclusion

This work demonstrates that the inherent non-linearity of phase-contrast HRTEM can be leveraged rather than just tolerated. By explicitly modeling the physics of multiple scattering, the multislice reconstruction method enables AET on thicker, more beam-sensitive materials than previously possible.

While this study utilized highly realistic simulations, future work will focus on addressing experimental challenges such as sample drift and precise tilt-axis calibration. The authors suggest that this framework could eventually be extended to biological cryo-electron microscopy (cryo-EM) for large protein complexes where standard linear approximations begin to fail. Ultimately, this method provides a robust path toward mapping the 3D atomic structure of complex, disordered, and dose-sensitive materials.

Summary updated

Abstract

Electron tomography is a technique used in both materials science and\nstructural biology to image features well below optical resolution limit. In\nthis work, we present a new algorithm for reconstructing the\nthree-dimensional(3D) electrostatic potential of a sample at atomic resolution\nfrom phase contrast imaging using high-resolution transmission electron\nmicroscopy. Our method accounts for dynamical and strong phase scattering,\nproviding more accurate results with much lower electron doses than those\ncurrent atomic electron tomography experiments. We test our algorithm using\nsimulated images of a synthetic needle geometry dataset composed of an\namorphous silicon dioxide shell around a silicon core. Our results show that,\nfor a wide range of experimental parameters, we can accurately determine both\natomic positions and species, and also identify vacancies even for light\nelements such as silicon and disordered materials such as amorphous silicon\ndioxide and also identify vacancies.\n
BibTeX
@misc{ren2018practical,
  author = {David Ren and Michael Chen and Laura Waller and Colin Ophus},
  title = {{A Practical Reconstruction Method for Three-Dimensional Phase Contrast\textbackslash\{\}n Atomic Electron Tomography}},
  year = {2018},
  eprint = {1807.03886},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1807.03886},
}
Figure