Machine learning improves X-ray imaging of nanotextures
(Nanowerk News) Using a combination of high-power X-rays, phase capture algorithms, and machine learning, Cornell researchers reveal complex nanotextures in thin film materials, offering scientists a new, simplified approach to analyzing potential candidates for quantum computing and microelectronics, among other applications other.
Scientists are especially interested in nanotextures that are distributed non-uniformly across thin films because they can impart new properties to materials. The most effective way to study nanotextures is to visualize them directly, a challenge that usually requires complex electron microscopy and does not preserve samples.
The new imaging technique is detailed in Proceedings of the National Academy of Sciences (“Real space imaging of periodic nanotextures in thin films via phasing diffraction data”) overcomes this challenge by using phase capture and machine learning to turn conventionally collected X-ray diffraction data – such as that generated at the Cornell High Energy Synchrotron Source, where data for the study is collected – into real-space visualizations of matter at the nanoscale.
The use of X-ray diffraction makes the technique more accessible to scientists and allows for imaging of larger portions of the sample, said Andrej Singer, assistant professor of materials science and engineering and David Croll Sesquicentennial Faculty Fellow in Cornell Engineering, who led the research. with doctoral student Ziming Shao.
“Large area imaging is important because it represents the true state of matter,” Singer said. “Nanotextures measured by local probes can depend on the choice of sites examined.”
Another advantage of the new method is that it doesn’t require the sample to be broken up, allowing for dynamic studies of thin films, such as passing in light to see how structures evolve.
“This method can be easily applied to study in-situ or operando dynamics,” said Shao. “For example, we plan to use methods to study how structures change in picoseconds after excitation with short laser pulses, which might enable new concepts for terahertz technology in the future.”
This technique was tested on two thin films, the first having a known nano-texture used to validate the imaging results. After testing a second thin film – a Mott insulator with physics related to superconductivity – the researchers discovered a new type of morphology that had never been observed in materials before – a strain-induced nanopattern that forms spontaneously upon cooling to cryogenic temperatures.
“Images extracted without prior knowledge,” said Shao, “have the potential to set new benchmarks and inform new physical hypotheses in phase field modeling, molecular dynamics simulations, and quantum mechanical calculations.”
