Analysis of high dimensional properties in quasicrystals with machine-learning molecular simulations

Nagai, Yuki   ; Iwasaki, Yutaka*; Kitahara, Koichi*; Kimura, Kaoru*; Shiga, Motoyuki   

In this talk, I give an overview of machine learning molecular simulation, a field that has recently been very active in research, and introduce its application to the analysis of high dimensionality in quasicrystals. I introduce machine learning molecular simulations in an easy-to-understand manner for those outside the field as a useful example of machine learning applications to physics. A quasicrystal is a material that has an ordered crystal structure but is not periodic. The crystal structure of quasicrystals can be understood as the projection of a superlattice in higher dimensional (5D or 6D) space onto real space (3D). In this talk, I report the results of specific heat analysis using experiments and machine learning molecular simulations to show that this higher dimensionality is actually reflected in observable physical quantities in the real world.