Increased funding for digital economy initiatives in Australia – OpenGov Asia
With a resolution 1000 times greater than that of an optical microscope, electron microscopes are exceptionally good at imaging materials and detailing their properties. But like all technologies, they have certain limitations.
To overcome these limitations, scientists have traditionally focused on upgrading hardware, which is expensive. But researchers at the US Department of Energy’s (DOE) Argonne National Laboratory show that advanced software developments can improve performance.
Our method helps improve the resolution of existing instruments so that users don’t need to upgrade to expensive new hardware so often.
– Tao Zhou, scientific assistant, author Argonne & Lad
Researchers at Argonne recently discovered a way to improve the resolution and sensitivity of an electron microscope using only an artificial intelligence (AI) framework. Their approach, published in npj Computational Materials, allows scientists to gain even more detailed information about the materials and the microscope itself, which may further expand its uses.
Electrons act like waves when they travel, and electron microscopes use this knowledge to create images. Images are formed when a material is exposed to a beam of electronic waves. By the way, these waves interact with the material, and this interaction is picked up by a detector and measured. These measurements are used to construct a magnified image.
In addition to creating magnified images, electron microscopes also capture information about the properties of materials, such as magnetization and electrostatic potential, which is the energy required to move a charge against an electric field. This information is stored in a property of the electronic wave called phase. Phase describes the location or timing of a point in a wave cycle, such as the point where a wave reaches its peak.
When measurements are taken, phase information is apparently lost. As a result, scientists cannot access information about magnetization or electrostatic potential from the images they acquire. Knowing these characteristics is essential for controlling and designing the desired properties in materials for batteries, electronics and other devices. Therefore, retrieving phase information is important.
Retrieving phase information is a decades-old problem. It comes from X-ray imaging and is now shared by other fields, including electron microscopy. To solve this problem, the computer scientists of Argonne propose to exploit tools designed to train deep neural networks, a form of AI.
Neural networks are basically a series of algorithms designed to mimic the human brain and nervous system. When given a series of inputs and outputs, these algorithms seek to plot the relationship between the two. But to do this accurately, neural networks must be trained. This is where the training algorithms come in.
Using these training algorithms, the research team demonstrated a way to retrieve phase information. But what makes their approach unique is that it also allows scientists to retrieve essential information from their electron microscope.
Their method also improves the resolution and sensitivity of existing equipment. This means that researchers will be able to recover tiny phase changes and, in turn, gain insight into small changes in magnetization and electrostatic potential, all without the need for expensive hardware upgrades.
As reported by OpenGov Asia, DOE’s Argonne National Laboratory has received nearly $ 3 million in funding for two interdisciplinary projects that will further develop artificial intelligence (AI) and machine learning technology.
Both grants were presented by the Office of Advanced Scientific Computing Research (ASCR) at DOE. They will help Argonne scientists and collaborators seek AI and machine learning work in developing approaches to manage huge datasets or develop better results where minimal data exists.
By integrating mathematics and scientific principles, they will build strong and accurate surrogate models. These types of models can significantly reduce the time and cost of complex simulations, such as those used to forecast climate or weather.