Machine learning method to exploit large volumes of X-ray data will accelerate material discovery.
Color coding makes aerial maps much easier to understand. Thanks to color, we can tell at a glance where a road, a forest, a desert, a city, a river or a lake is.
In collaboration with several universities, the U.S. Department of Energy’s (DOE) Argonne National Laboratory has pioneered a method for creating color-coded graphs of large volumes of data from X-ray scans. This new tool uses computer sorting of data to find clusters related to physical properties, such as atomic distortion in a crystal composition. It should dramatically accelerate future research into atomic-scale structural changes induced by temperature variants.
“Our method uses machine learning to rapidly analyze huge amounts of data from X-ray diffraction,” said Raymond Osborn, senior physicist in Argonne’s Materials Science Division. “What might have taken us months in the past, now takes approximately a quarter of an hour, with much finer results.”
Since more than one century, X-ray diffraction, or XRD, is one of the most successful scientific methods for analyzing materials. It has provided key insights into the 3D atomic composition of countless technologically important materials.
Over the past few decades. a DOE Business of Science user facility in Argonne. However, analysis methods that can cope with these huge datasets are sorely lacking.
The team calls its new method X-ray Temperature Clustering, or XTEC for short . It accelerates material discovery through rapid clustering and color-coding of large x-ray datasets to reveal previously hidden structural changes that occur as temperature rises or falls. A typical large dataset would be 10 000 gigabytes, which equals approximately 3 tens of millions of songs streaming music.
XTEC relies on the power of machine learning not supervised, using methods developed for this project at Cornell University. This machine finding out does not depend on initial training and learning with already well-studied data. Instead, it learns by finding patterns and clusters in large datasets without such training. These patterns are then represented by a color code.
“For example, XTEC could assign red to data cluster one, which is associated with a certain property that adjusts with temperature in a particular way,” Osborn said. “Then group two would be blue and associated with another property with a different temperature dependence, and so on. The colors indicate whether each group represents the equivalent of a road, a forest or a lake on an aerial map.”
As a test, XTEC has analyzed data from the APS 6-ID-D beamline, extracted from two superconducting crystalline materials at temperatures near absolute zero. At this ultra-low temperature, these materials go into a superconducting state, offering no resistance to electric current. More important for this study, other unusual features emerge at higher temperatures related to changes in the structure of the material.
By applying XTEC, the team extracted a unprecedented amount of information about atomic framework changes at different temperatures. These include not only distortions in the ordered arrangement of atoms in the material, but also fluctuations that occur when such changes occur.
“Thanks to the machine learning, we are able to see the behavior of materials not noticeable by conventional XRD,” Osborn said. “And our method is applicable to many big data problems not only in superconductors, but also in batteries, solar cells, and any temperature-smart device.”
The APS is undergoing a significant upgrade that will increase the brightness of its X-ray beams up to 500 time. The upgrade will come with a significant increase in data collected at APS, and machine learning techniques will be essential to analyze this data in a timely manner.
In addition to Osborn, Argonne writers include Matthew Krogstad, Daniel Phelan, Puspa Upreti, Michael Norman, and Stephan Rosenkranz. The primary collaborating partner is Cornell University (Eun-Ah Kim, Jordan Venderley, Krishnanand Mallayya, Michael Matty, Geoff Pleiss, Varsha Kishore, and Kilian Weinberger) and Cornell Substantial Strength Synchrotron Supply (Jacob Ruff). Other partners include the University of Tennessee (David Mandrus), University of Maryland (Lekh Poudel) and New York University (Andrew Gordon Wilson).
Funding for Argonne was provided by the DOE’s Office of Basic Energy Sciences and the Nationwide Science Foundation.