An atomic-resolution snapshot of a system can be obtained if researchers shoot a pair of X-ray pulse just moments apart. Future generations of super-fast computers, communications, and other technologies could be improved by comparing these snapshots.
Artificial intelligence was used to automate the process of resolving the X-ray snapshots. The machine learning-aided method was published in Structural Dynamics on October 17th.
Turner said that they can now access a different range of measurements.
The blob needs to be handled.
The two-pulse technique uses X-rays to scatter off a material and then to detect one photon at a time. A blotchy image that represents the precise configuration of the sample at one instant in time is produced by a detector thatmeasures the scattered photons. The researchers look at the speckle patterns from each pulse to calculate fluctuations.
Turner said that every photon causes an explosion of electrical charge on the detector. The charge clouds merging together to create a blob. The researchers need tons of data to understand the speckle pattern.
"You need a lot of data to figure out what's going on in the system," said Sathya Chitturi, a PhD student who led this work. Turner and Mike Dunne are the directors of the Linac Coherent Light Source.
All of the data had to be collected first, then analyzed using models that estimate how the photons bunch together at the detector, taking a long time to understand the speckle patterns.
The raw detector image of scattered photons is used in the machine learning method. The method is ten times faster on its own and 100 times faster with improved hardware.
Part of the success of the new method was due to the work of Nicolas Burdet, an associate staff scientist at SLAC. Through this training, the program was able to learn how the charge clouds form, and how many photons hit the detector. Under very blobby conditions, the model was accurate.
Looking past the clouds.
High- temperature superconductors or quantum spin liquids can be difficult to study because X-rays scatter off them too weakly. The new method can be applied to other non-quantum materials.
Turner believes that the research will help the LCLS- II upgrade, which will allow researchers to collect up to a million images, or a few terabytes of data, per second, compared to about a hundred images a second for LCLS.
Turner said that they were excited about the upgrade but worried if they could handle it. The team's new technique should be fast enough to deal with all the data. The new formula will help.
Artificial intelligence promises to change the process itself. Researchers will be able to analyze data and make changes during the experiment if they don't have to make decisions after the data is collected. Researchers will be able to investigate unexpected phenomena in real time thanks to it.
"This method allows you to explore more of the materials science you're interested in and maximize scientific impact by allowing you to make decisions at different points along your experiment about changes in experimental variables such as temperature, magnetic field, and material composition."
Machine learning is being used to advance materials and chemistry research.
There is more information about Sathya R. Chitturi et al. There is a book called 10.1063/4.0000161.
In preparation for high data rate X-ray facilities, Hongwei Chen and his co-workers tested the data framework for an artificial intelligence program. There is a book titled "arxiv.2210.10137."
Journal information: arXiv
All Rights Reserved © 2024
