Researchers apply deep learning to X-ray diffraction

Discover hidden mysteries in nature: Researchers practice deep learning on X-ray emissions

The sharp increase in the intensity of the reflected light for a certain angle of incidence is represented as a peak, and its pattern can be used as a “fingerprint / characteristic of a compound”. A top model has several peaks corresponding to one component. The test data emission model (red curve) is covered by the results of the automatic encoder (blue curve), which represents the relevance of the peaks in the characterization. The apex labeled (a) has a noticeable intensity but low relevance as a result of in-depth study. Credit: Ryo Maezono from JAIST

X-ray diffraction (XRD) is an experimental technique to find out the atomic structure of an object by irradiating it with X-rays at different angles. Importantly, the intensity of the reflected X-rays rises at a certain radiation angle, forming a pattern of the apex of the diffusion. XRD serves as a fingerprint for the material as each substance creates a unique pattern.

In research and development, changes in XRDs are used to determine the position and quantity of additional elements that need to be added to modify the material to help increase the desired functionality, say, the efficiency of energy storage in the battery.

However, the peak changes in XRDs are almost incomprehensible to humans. This makes specifying the features and relevance of the various peaks difficult for material identification. To this end, a group of Japanese researchers led by Professor Ryo Maezono from the Japan Institute of Advanced Science and Technology (JAIST) have applied a Deep Learning technique called “automatic encoding” to problems to find hidden accuracy in In XRDs that can accelerate the development of new functional materials.

The research team also included Associate Professor Kenta Hongo and Assistant Professor Kousuke Nakano from JAIST. Their work is published in Advanced Theory and Simulation.

Explaining the basics of auto-encoder technology, Professor Maezono says, “Auto-encoder technology captures data features by displaying them as dots on a feature space. Grain information. Compact automatic encoder “The data size will be able to capture multi-dimensional XRD analysis effectively in two-dimensional planes.”

Using neural networks, the researchers applied an automatic codec to 150 XRD models of alloys of different concentrations. In the interval, each XRD feature is projected to a single point. These points form clusters in which similar elements with similar concentrations of elements are placed close together. Thus, the distance between the points in the feature space allows an estimate of the concentration of any given model alloy wheels. This also allows for indirect alloy modification by identifying the XRD peak that changes when a new element is added to the alloy or its component ratio is changed.

Researchers have further proposed a novel program of featured places. When the highest interest is masked on the original XRD model, the dots on the functional space change. The extent of the transition helps to distinguish whether the relevance of the top is to capture the properties of the material. Using this technique, researchers can identify which peaks are relevant to be careful about estimating the amount of doping material. Learning.

Researchers have also proposed a program of automated coding for the creation of artificial XRD models by incorporating existing objects to address small changes in metallic composition. This method will create a reliable data set, avoiding expensive ab initio simulations to calculate.

“The results of this research are not limited to the top XRD models. Instead, they provide general deep learning techniques that can be used to extract features from material science data. Its framework can detect accuracy. Hidden in nature, people can not identify and are expected to serve as powerful tools for theorem discovery through data science, “said Professor Maezono.

The application of the described automation codecs can accelerate the development of high efficiency, low cost and low environmentally friendly materials, starting in the new era of Deep Science based material science research.

Additional information:
Keishu Utimula et al, XRD Model Spacing Features Built by Autoencoder, Advanced Theory and Simulation (2022). DOI: 10.1002 / adts.202200613

Provided by Japan Institute of Modern Science and Technology

Excerpt: The Search for Hidden Regularity in Nature: Researchers Practice Deep Learning on X-Ray Radiation (December 2222, 22 December) Retrieved December 22, 2022 from / 2022-12-hidden-regularities-nature-deep-x -ray.html

This document is subject to copyright. Except for fair settlement for the purpose of private study or research, no part may be reproduced without written permission. Content is provided for informational purposes only.


Also Read :  Michael Flynn Claims People Working on Putting Robotics in Us, Changing DNA

Leave a Reply

Your email address will not be published.