Scanning probe microscopy (SPM) has revolutionized materials science, nanotechnology, chemistry, and biology by making it possible to map surface properties and manipulate surfaces with atomic precision. These accomplishments, though, still require ongoing human supervision; completely automated SPM is still a work in progress.
Researchers present a machine learning-based artificial intelligence system for autonomous SPM operation. (DeepSPM). Convolutional neural networks are used in DeepSPM to evaluate the quality of acquired images, a deep reinforcement learning agent reliably conditions the state of the probe and an algorithm for finding suitable sample regions.
DeepSPM can constantly collect and categorize data in multi-day scanning tunneling microscopy experiments while managing the probe quality in response to changing experimental conditions. Their strategy paves the way for cutting-edge techniques that are hardly humanly possible. (e.g., large dataset acquisition and SPM-based nanolithography).
A physical quantity, such as force in atomic force microscopy (AFM) or quantum tunneling current in scanning tunneling microscopy (STM), is measured as a function of probe position by scanning an atomically sharp probe near (typically 1 nm) above a surface. It makes it possible to create an image of the scanned area. This technique enables measurements of various materials in various conditions, from ambient to ultra-high vacuum at cryogenic temperatures.
Related Content: Cryo Electron Microscopy Visualizes Coronavirus Structure