With resolution 1000 times greater than a light microscope, electron microscopes are exceptionally good at imaging materials and detailing their properties. Researchers at the U.S. Department of Energy’s (DOE) Argonne National Laboratory are demonstrating that advanced software developments and an artificial intelligence (AI) framework can push their performance even further.

Argonne’s Charudatta Phatak views a magnified image from a Lorentz transmission electron microscope. Phatak’s team is using AI to improve microscope sensitivity and accuracy. (Photo: Argonne National Laboratory)

Along with creating magnified images, electron microscopy techniques also capture information about material properties, such as magnetization and electrostatic potential, which is the energy needed to move a charge against an electric field. The information is stored in a property of the electron wave known as phase. Phase describes the location or timing of a point within a wave cycle, such as the point where a wave reaches its peak.

Retrieving phase data, however, is a decades-old problem for scientists. Information such as magnetization and electric potentials is carried by the phase of the electron wave and is lost during the image acquisition process.

To allow scientists to access such data, the Argonne researchers proposed leveraging tools built to train deep neural networks, a form of artificial intelligence (AI) that mimics the human brain and requires training algorithms.

With training data, the Argonne research team demonstrated a way to recover phase information, as well as essential information about their electron microscope, including spatial resolution, accuracy, and sensitivity of the microscopy. The AI-enabled analysis of high-resolution images – a process known as “reverse-mode automatic differentiation” – determines atomic positions to infer physical properties about the materials.

Researchers recover tiny shifts in phase, and in turn, receive information about small changes in magnetization and electrostatic potential, all without requiring costly hardware upgrades.

Read the report  .

For more information, contact Argonne National Laboratory; 630-252-2000.