Innovative Deep Learning Model at Tokyo University Accelerates Phase Identification in Multiphase Crystalline Materials

November 17, 2023

Innovative Deep Learning Model at Tokyo University Accelerates Phase Identification in Multiphase Crystalline Materials

Science / Artificial Intelligence & Machine Learning

A groundbreaking study led by Junior Associate Professor Tsunetomo Yamada from Tokyo University of Science, in collaboration with several prominent Japanese research institutions, has marked a significant advancement in the field of materials science!

“Across the world, researchers have made attempts to predict new substances using artificial intelligence and machine learning. However, identifying whether a desired substance is produced takes up substantial time and effort on the part of human experts. Therefore, we came up with the idea of using deep learning to identify new phases,” – Dr. Yamada.

Moreover, the research team developed a deep learning model capable of identifying previously unknown quasicrystalline phases in multiphase crystalline samples, a breakthrough published in the journal Advanced Science on November 14, 2023.

The Challenge in Identifying Crystalline Phases

Crystalline materials, with their ordered atomic structures, are crucial in various applications, including semiconductors, pharmaceuticals, and photovoltaics. Traditional methods like powder X-ray diffraction have been instrumental in identifying these structures. However, analyzing multiphase samples, which contain a mix of different crystals, poses significant challenges. The conventional process is not only complex but also time-consuming, heavily relying on the expertise of scientists.

A Leap Forward with Deep Learning

Addressing this challenge, Dr. Yamada and his team have harnessed the power of machine learning to create a “binary classifier” model. This model is specifically designed to detect icosahedral quasicrystal (i-QC) phases, known for their unique long-range order and self-similarity in diffraction patterns. The development of this model involved using convolutional neural networks, training it with synthetic multiphase X-ray diffraction patterns representing i-QC phases.

Impressive Accuracy and Applications

The model’s effectiveness is underscored by its remarkable prediction accuracy of over 92%. It successfully identified an unknown i-QC phase in a multiphase Al–Si–Ru alloy sample, further validated through transmission electron microscopy. This achievement is significant, as it demonstrates the model’s capability to detect quasicrystalline phases even when they are not the dominant component in the mixture.

Potential for Wider Applications

The implications of this deep learning model extend beyond i-QC phases. It holds promise for identifying new decagonal and dodecagonal quasicrystals and can potentially be adapted to various other crystalline materials. This versatility points to a broader impact in materials science, particularly in the discovery and analysis of novel materials.

A Breakthrough in Materials Science

The success of this deep learning model in rapidly identifying new phases in quasicrystals represents a major step forward in the field. Not only accelerating the process of phase identification in multiphase samples! However, also opens up new possibilities for the discovery of unique materials. As a result, these materials could play pivotal roles in addressing emerging challenges in energy storage, carbon capture, and advanced electronics.

Conclusion and Future Outlook

The innovative approach taken by Dr. Yamada and his team showcases the transformative potential of integrating machine learning with materials science. As the team looks to the future, there is optimism that this model will lead to further breakthroughs in the field. The study stands as a testament to the power of interdisciplinary collaboration in pushing the boundaries of scientific discovery.