New AI-Analysis for Materials
FHI researchers have developed a new analysis method "ARISE" for materials based on the use of artificial intelligence. Today, the research report by Andreas Leitherer, Angelo Ziletti and Luca M. Ghiringhelli is published in the journal Nature Communications.
"ARISE is a big step forward," says physicist Andreas Leitherer, explaining the results of his four-year doctoral research at the NOMAD Laboratory and within the BiGmax research network. The new method for identifying crystal structures, which is based on the AI method "Bayesian Deep Learning," can be used to study more than a hundred different materials. Because of the program's learning ability, it can also recognize irregular single- and polycrystalline systems, from both synthetic and experimental sources. For example, it has been used in electron tomography experiments of metallic nanoparticles.
Furthermore, defects such as grain boundaries can be characterized, which is of relevance, for example, to industrial steel, the use of which is firmly embedded in our everyday lives. Interesting is also the use in two-dimensional materials like graphene or one-dimensional systems like carbon nanotubes. The program, which can be learned and expanded, can be used to recognize previously hidden patterns in databases, promising digital storage media and also in the production of solar cells," explains Andreas Leitherer. ARISE is to be further developed in the coming years.