Specific heat anomalies and local symmetry breaking in (anti-)fluorite materials; A Machine learning molecular dynamics study

Kobayashi, Keita ; Nakamura, Hiroki  ; Okumura, Masahiko   ; Itakura, Mitsuhiro  ; Machida, Masahiko  

The specific heat anomaly in (anti-)fluorite structures was analyzed using machine learning molecular dynamics (MLMD) methods. By employing the Farthest Point Sampling method and the Bootstrap method, first-principles training data were efficiently generated, and machine learning potentials were created for thorium dioxide (fluorite structure) and lithium oxide (anti-fluorite structure). As a result, the MLMD method accurately reproduced the reported thermal properties of thorium dioxide and lithium oxide. These materials exhibit a specific heat anomaly at high temperatures due to sublattice disordering. However, the details are complex and not fully understood. In this study, by applying a local order parameter methodology, which has been used in the analysis of liquid-liquid phase transitions, we revealed that the anomalous specific heat in (anti-)fluorite structures can be interpreted as a consequence of local symmetry breaking.