Atomic thermal fluctuation reduction method for robust local lattice structure identification in finite-temperature molecular dynamics

Classical molecular dynamics (MD) is extensively employed to explore the properties, deformations, and fractures of materials at the atomic scale. Identifying local structures is crucial for understanding the mechanisms behind material deformation and fracture. Nevertheless, analyzing the local lattice structure at high temperatures poses challenges due to atomic thermal fluctuations, which act as noise and potentially lead to misjudgment of the local lattice structure. To date, various strategies have been implemented to circumvent this issue. However, they cannot be a solution because it is unable to reproduce phenomena unique to high temperatures, whereas others require significant computational resources. This paper introduces an innovative method to reduce atomic thermal fluctuations using a straightforward algorithm, thereby facilitating accurate identification of local lattice structures even at high temperatures. Our approach incorporates novel degrees of freedom, termed ‘Markers,’ that are linked to atoms. By reducing the thermal fluctuation of these Markers, precise analysis of the local lattice structure becomes feasible. The efficacy of this method is validated through its thermal reducibility and Markers trackabilities to atoms. Utilizing common neighbor analysis, the error rate for structure identification with our method is nearly 0% at temperatures up to 1200 K in Fe, in contrast to approximately 5% without it. Furthermore, the average distance between atoms and Markers remains below 0.1 Å. Applying our method to phase transformations, we successfully observed the transition from face-centered cubic to body-centered cubic structure in Fe at 1200 K. This method holds promise for expanding the capabilities of MD simulations at high temperatures.