X-rays can be used as cameras with fast atomic resolution, and if researchers take a couple of X-rays apart for a while, they will get a photo of the atomic resolution of a system in two moments. These photo comparisons show how material changes in a matter of seconds, helping scientists design computers, communications, and other high-speed technologies.
However, resolving the information in these X-ray images is difficult and time consuming, so Joshua Turner, a leading scientist in the Department of Energy at the National Acceleration Center, SLAC and Stanford University, and ten researchers Others have turned to artificial intelligence to automate the process. . Methods to Assist Their Machine Learning Published on October 17th Structural dynamicsAccelerate this X-ray probe technique and extend it to previously inaccessible material.
“The most exciting thing for me is that we can now access different ranges of measurement that we could not have done before,” Turner said.
When studying materials using these two pulse techniques, X-rays scatter from one object and usually detect one photon at a time. These scattering photon sensors are used to create the scatter pattern – a cracked image that represents the precise configuration of the model at the same time. The researchers compared the samples of the nodules from each pair to calculate the sensitivity of the samples.
“However, every photon generates an explosion of electric charge on the detector,” Turner said. “If there are too many photons, these charge clouds combine to form an unrecognizable block.” This cloud of noise means that researchers have to collect tons of scattered data to provide a better understanding of Scratch pattern.
“You need a lot of data to understand what is going on in the system,” said Dr. Sathya Chitturi. Stanford University students lead this work. He was introduced by Turner and author Mike Dunne, director of the Linac Coherent Light Source (LCLS) X-ray laser at SLAC.
With the simple method, all data must be collected first, then analyzed using a model that estimates how the photons gather at the sensor – a long process to understand the pattern of light.
On the other hand, machine learning methods use raw image capture of scattering photons to extract directly adaptive information. The new method is 10 times faster on its own and 100 times faster when combined with improved hardware that allows data analysis closer to real-time.
Part of the success of this new method comes from the efforts of author Nicolas Burdet, a science staff member at SLAC, who developed a simulation machine that produces data for machine learning training. Through this training, the algorithm can learn how charge clouds merge and solve the number of photons that hit the detector per block and per pair of pulses. This model proves accurate even under extremely disturbing conditions.
Visibility beyond the clouds
This model can extract information for many materials that are difficult to study because the X-rays scattered from them are too weak for detection, such as high-temperature storage devices or liquids orbiting the quantum. Chituri said the new method could also be applied to other non-Quantum materials, including colloids, alloys and glasses.
Turner said the research should help with the LCLS-II upgrade, which will allow researchers to capture up to a million images or a few terabytes of data per second, compared to about a hundred images per second. LCLS.
“At SLAC, we are excited about this upgrade, but also worried if we can manage this amount of data,” Turner said. In the relevant documents, the team found that their new technique should be fast enough to deal with all that data. “This new algorithm will definitely help.”
The acceleration provided by artificial intelligence promises to change the experimental process itself as well. Instead of making decisions after data collection and analysis, researchers will be able to analyze data and make changes during data collection, which can save time and money spent during the experiment. It will also allow researchers to detect surprises and redirect their experiments in real time to investigate unexpected phenomena.
“This method allows you to learn more about the material science you are interested in and enhances your scientific influence by allowing you to make decisions at different points based on your experience on changes in experimental variables such as temperature, magnetic field and material composition,” said Chitturi. That
The study is part of a larger collaboration between SLAC, Northeastern University, and Howard University to use machine learning to advance materials and chemistry research.
Sathya R. Chitturi et al, learning machine photon detection algorithm for integrated ultraviolet sensitivity analysis, Structural dynamics (2022). DOI: 10.1063 / 4.0000161
Hongwei Chen et al, Data framework testing for AI algorithms in preparation for high-rate X-ray equipment arXiv (2022). DOI: 10.48550 / arxiv.2210.10137
Provided by SLAC National Accelerator Laboratory
Excerpt: Artificial Intelligence ‘Cloud’ Sensor to Accelerate Material Research (November 7, 2022, November 7, 2022) Retrieved November 7, 2022 from https://phys.org/news/2022-11-artificial-intelligence -deciphers-detector-clouds.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.