Crystalline materials are characterized by their ordered, three-dimensional structures composed of atoms, ions, or molecules. Widely employed in various fields, including semiconductors, pharmaceuticals, photovoltaics, and catalysts, these materials play a pivotal role in technological advancements. The exploration of crystalline structures has expanded to address emerging challenges in energy storage, carbon capture, and advanced electronics.
The identification of crystalline materials is crucial for their development. Currently, powder X-ray diffraction is a widely used method for this purpose, examining scattered X-rays from powdered samples to determine their structure. However, when dealing with multiphase samples containing different types of crystals with distinct structures, orientations, or compositions, the identification process becomes complex. Expertise is required to accurately identify various phases, making the process time-consuming.
To overcome these challenges, researchers have turned to innovative data-driven methods, particularly machine learning, to distinguish individual phases within multiphase samples. While progress has been made in using machine learning for known phases, the identification of unknown phases in multiphase samples remains a challenge.
In a recent breakthrough, a team of researchers from Tokyo University of Science, National Defense Academy, National Institute for Materials Science, Tohoku University, and The Institute of Statistical Mathematics proposed a new machine learning “binary classifier” model. This model aims to identify the presence of icosahedral quasicrystal (i-QC) phases—a type of long-range ordered solids with self-similarity in their diffraction patterns—from multiphase powder X-ray diffraction patterns.
Led by Junior Associate Professor Tsunetomo Yamada from Tokyo University of Science, the study involved creating a “binary classifier” using 80 types of convolutional neural networks. The researchers trained the classifier model using synthetic multiphase X-ray diffraction patterns designed to represent the expected patterns associated with i-QC phases. The model's performance was then assessed using both synthetic patterns and a database of actual patterns.
Remarkably, the model achieved a prediction accuracy of over 92%. It successfully identified an unknown i-QC phase within multiphase Al–Si–Ru alloys, screening 440 measured diffraction patterns from unknown materials in six different alloy systems. The presence of the unknown i-QC phase was further confirmed through analysis of the material's microstructure and composition using transmission electron microscopy.
The proposed deep learning method demonstrated the ability to identify the i-QC phase even when it is not the most prominent component in the mixture. Moreover, the model's application can extend to the identification of new decagonal and dodecagonal quasicrystals, showcasing its versatility across various types of crystalline materials.
Dr. Yamada highlighted the significance of the proposed model in detecting unknown quasicrystalline phases within multiphase samples with high accuracy. The success of this deep learning model suggests the potential for expediting the process of phase identification in multiphase samples, marking a breakthrough in the field of materials science.
In summary, this study represents a significant advancement in the identification of entirely new phases in quasicrystals, commonly found in materials such as mesoporous silica, minerals, alloys, and liquid crystals. The integration of machine learning techniques opens up possibilities for accelerated discovery and innovation in the realm of crystalline materials.
Source: Tokyo University of Science