As materials are subjected to pressure and relaxation, their properties undergo significant transformations.
As individual components, supplies undergo a process of continuous evolution and refinement. When subjected to pressure and relaxation, they exhibit distinct behavioural patterns. Researchers have devised a groundbreaking method combining X-ray photon correlation spectroscopy (XPCS), artificial intelligence (AI), and machine learning to quantify the dynamics of supply fluctuations.
This approach generates distinct signatures for diverse materials, which are then accessible for analysis by a neural network to produce novel insights inaccessible to researchers prior to this development. A neural network is a type of computer model that processes information similarly to how the human brain functions.
Researchers at the Advanced Photon Source (APS) and Center for Nanoscale Materials (CNM) of the U.S. Scientists at the Division of Vitality’s Argonne Nationwide Laboratory have successfully combined X-ray photon correlation spectroscopy (XPCS) with an unsupervised machine learning algorithm, specifically a self-taught artificial neural network that doesn’t require human supervision. The algorithm learns to recognize patterns within the scattering of X-rays by a colloidal suspension, comprising a multitude of particles dispersed in solution. The Advanced Photon Source (APS) and Center for Nanoscale Microscopy (CNM) serve as consumer-facing amenities within the Department of Energy’s Office of Science.
“The primary objective of the AI is to analyze scattering patterns by treating them as ordinary photographs or video footage, then identifying recurring patterns through data digestion.” The artificial intelligence serves as a proficient tool in recognizing samples.
“According to Argonne postdoctoral researcher James ‘Jay’ Horwath, lead author of the study, understanding how supplies change and evolve over time is achieved by compiling X-ray scattering data.”
These complex patterns prove challenging for scientists to identify without the aid of AI-powered tools. “As the X-ray beam is trained on the complex patterns, even experts find themselves struggling to decipher their meaning due to the sheer volume and intricacy of the data.”
To enable researchers to better grasp their findings, they must distill complex data into concise summaries that convey the most salient insights regarding the pattern. According to Horwath, considering a fabric’s digital twin is akin to possessing its genetic code, encompassing all the information necessary to regenerate every aspect of the material.
The venture is called Synthetic Intelligence for Non-Equilibrium Leisure Dynamics, abbreviated as AI-NERD. Fingerprints are generated using an innovative technique called an autoencoder. An autoencoder is a type of neural network that compresses distinct visual data into a latent representation, often referred to as a “fingerprint” among experts, which also incorporates a decoder mechanism that enables the reconstruction of the original image from this compacted latent space.
Researchers sought to develop a comprehensive mapping of fabric fingerprint patterns, grouping similar characteristics together in distinct clusters. Through a holistic examination of the diverse fingerprint neighborhood patterns on the map, the researchers have gained insight into the spatial organization and evolution of these features over time, as they were shaped and reshaped by various pressures and relaxation processes.
Artificial intelligence, boasting impressive fundamental sample recognition abilities, is well-equipped to accurately categorise diverse X-ray images, thereby creating a precise mapping system. “The primary objective of the AI system is to analyze scattering patterns by treating them as conventional images or video feeds, subsequently identifying recurring patterns through a process of digestion,” Horwath explained. “The artificial intelligence is a proficient expert in sample recognition.”
As the advanced Particle Spectrometer goes live, leveraging AI-driven insights into scattering patterns will be crucial. The new facility will produce X-ray beams that are a staggering 500 times brighter than those of the existing Advanced Photon Source (APS). “When we receive information from the upgraded Advanced Planning System, we’ll need to utilize artificial intelligence capabilities to process it effectively,” Horwath stated.
Researchers from the Idea Group at the Center for Nanoscale Materials (CNM) joined forces with experts from Argonne’s X-ray Science division to conduct advanced molecular simulations, mirroring experimental findings in polymer dynamics observed through X-ray Photon Correlation Spectroscopy (XPCS). The goal was to generate synthetic data to inform AI workflows and train the AI-NERD system.
The examination was supported through a laboratory-directed research and development grant from Argonne.
The authors of the examine are: James (Jay) Horwath, Xiao-Min Lin, Hongrui He, Qingteng Zhang, Eric Dufresne, Miaoqi Chu, Subramanian Sankaranaryanan, Wei Chen, Suresh Narayanan, and Mathew Cherukara from Argonne. Chen and He hold joint faculty appointments at the University of Chicago, while Sankaranarayanan has a joint affiliation with the University of Illinois at Chicago.
