Large-Angle Convergent-Beam Electron Diffraction Patterns via Conditional Generative Adversarial Networks

Abstract

We show how generative machine learning can be used for the rapid computation of strongly dynamical electron diffraction directly from crystal structures, specifically in large-angle convergent-beam electron diffraction (LACBED) patterns. We find that a conditional generative adversarial network can learn the connection between the projected potential from a cubic crystal's unit cell and the corresponding LACBED pattern. Our model can generate diffraction patterns on a GPU many orders of magnitude faster than existing direct simulation methods. Furthermore, our approach can accurately retrieve the projected potential from diffraction patterns, opening a new approach for the inverse problem of determining crystal structure.

9 pages, 7 figures

References in corpus (12)

• Auto-Encoding Variational Bayes
• Conditional Generative Adversarial Nets
• Solving the Quantum Many-Body Problem with Artificial Neural Networks
• A high-bias, low-variance introduction to Machine Learning for physicists
• Drawing Phase Diagrams of Random Quantum Systems by Deep Learning the Wave Functions
• Phase diagram study of a two-dimensional frustrated antiferromagnet via unsupervised machine learning
• 'Digital' Electron Diffraction - Seeing the Whole Picture
• Phase Object Reconstruction for 4D-STEM using Deep Learning
• The percolating cluster is invisible to image recognition with deep learning
• Structure refinement from 'Digital' Large Angle Convergent Beam Electron Diffraction Patterns
• Machine learning the 2D percolation model
• Solving the Crystallographic Phase Problem using Dynamical Scattering in Electron Diffraction