Reviewed by Lexie CornerJul 17 2024
Researchers from the Advanced Photon Source (APS) and Center for Nanoscale Materials (CNM) at the US Department of Energy's (DOE) Argonne National Laboratory have paired XPCS with an unsupervised machine learning algorithm, a type of neural network that does not require expert training. This news research is published in Nature Communications.
The AI-NERD model learns to produce a unique fingerprint for each sample of XPCS data. Mapping fingerprints from a large experimental dataset enables the identification of trends and repeating patterns, which aids in understanding how materials evolve. Image Credit: Argonne National Laboratory
Like humans, materials change throughout time and behave differently when stressed and relaxed. Scientists studying the dynamics of material change have created a novel approach that combines X-ray photon correlation spectroscopy (XPCS), artificial intelligence (AI), and machine learning.
This technology produces “fingerprints” of different materials that can be read and evaluated by a neural network, yielding fresh information that scientists previously could not access. A neural network is a computer model that functions similarly to the human brain.
The algorithm teaches itself to detect patterns concealed in configurations of X-rays dispersed by a colloid, which is a group of particles suspended in solution. The APS and CNM are DOE Office of Science User Facilities.
The way we understand how materials move and change over time is by collecting X-ray scattering data.
James Horwath, Study First Author and Postdoctoral Researcher, Argonne National Laboratory
These patterns are too intricate for scientists to discover without the use of artificial intelligence.
Horwath stated, “As we are shining the X-ray beam, the patterns are so diverse and so complicated that it becomes difficult even for experts to understand what any of them mean.”
Researchers must condense all of the data into fingerprints that contain just the most crucial details about the sample to have a deeper understanding of the subject matter they are researching.
“You can think of it like having the material’s genome, it has all the information necessary to reconstruct the entire picture,” Horwath stated.
The project is called AI-NERD (Artificial Intelligence for Non-Equilibrium Relaxation Dynamics). The process of creating the fingerprints involves using an autoencoder. An autoencoder is a type of neural network that converts the original image data into a fingerprint, or what scientists refer to as a latent representation. It also has a decoder algorithm that allows it to convert the latent representation back into the original image.
The researchers wanted to map the fingerprints on the material by grouping fingerprints that shared similar features into neighborhoods. By taking a comprehensive look at the characteristics of the different fingerprint neighborhoods on the map, the researchers were able to gain a better understanding of the structure of the materials and how they changed over time as they underwent stress and relaxation.
Put simply, AI possesses strong general pattern recognition skills, which enable it to effectively classify and arrange the various X-ray images into the map.
“The goal of the AI is just to treat the scattering patterns as regular images or pictures and digest them to figure out what are the repeating patterns. The AI is a pattern recognition expert,” Howarth added.
Using AI to comprehend scattering data will be especially crucial once the improved APS goes online. The new facility will produce X-ray beams that are 500 times brighter than the original APS.
Howarth stated, “The data we get from the upgraded APS will need the power of AI to sort through it.”
The theory group at CNM cooperated with the computational group in Argonne’s X-ray Science division to conduct molecular simulations of the polymer dynamics exhibited by XPCS and, in the future, synthetically produce data for training AI processes such as the NERD.
An Argonne laboratory-directed research and development grant funded the study.
Argonne’s James (Jay) Horwath, Xiao-Min Lin, Hongrui He, Qingteng Zhang, Eric Dufresne, Miaoqi Chu, Subramanian Sankaranaryanan, Wei Chen, Suresh Narayanan and Mathew Cherukara are the study authors.
Chen and He hold joint appointments at the University of Chicago, while Sankaranaryanan holds a joint appointment at the University of Illinois Chicago.
Journal Reference:
Horwath, J. P., et al. (2024) AI-NERD: Elucidation of relaxation dynamics beyond equilibrium through AI-informed X-ray photon correlation spectroscopy. Nature Communications. doi:10.1038/s41467-024-49381-z.