In collaboration with a research group in the UK, we have developed CONQUEST, a large-scale first-principles calculation code. By employing a linear-scaling (order-N) method and an original localized-orbital approach, CONQUEST enables electronic-structure analysis, structural optimization, and molecular dynamics (MD) simulations based on the density functional theory for extremely large systems that cannot be treated using conventional methods.
The figure shows an example of research using CONQUEST: a theoretical analysis of a core–shell semiconductor nanowires based on large-scale first-principles calculations. By optimizing a structural model of a realistically sized nanowire, we have successfully calculated the strain distribution within the nanowire arising from the lattice mismatch between silicon in the core and germanium in the shell. We have also clarified the spatial distribution of the electronic states near the Fermi level, which are relevant to charge carriers. In January 2020, CONQUEST was released under the open-source MIT License, making the program freely available to researchers in Japan and worldwide.

Our group also applies large-scale first-principles MD simulations using CONQUEST to surfaces, defects, amorphous structures, and other complex systems. A major research challenge is to understand the diverse structures that emerge in molecular simulations of these systems and to clarify how they are related to the resulting physical properties. We have proposed a new method for analyzing the various local, medium-range structures observed in molecular simulations using unsupervised machine learning and have successfully demonstrated its ability. The figure on the right shows an analysis of the melt–quench process used to generate amorphous silicon structures. We are currently applying this method to a wide range of materials to further assess its effectiveness. In the future, we aim to extend the method and use it to design materials with superior functionalities of materials.