Cross-correlation for detecting and understanding patterns inab initiomolecular dynamics simulations of liquid metals

When analysing the motion of atoms during molecular dynamics simulations, it can be challenging to detect physically meaningful and reliable patterns. In the growing field of modelling liquid metal systems, detection of these patterns from trajectories has traditionally been limited to 'by-eye' recognition, which especially struggles to relate coupled motions that are time-lagged. In this work, we apply statistical cross-correlation techniques to relate time-series information about the positions, velocities, and energies of liquid metal atoms throughout a trajectory. Our analyses reveal previously unidentified temporal relationships between the motion of atoms in liquid Ga-In and Ga-Sn, and are able to link these patterns to trends in interaction energy with adsorbates. The methodology discussed here provides a powerful framework for analysing coupled atomic-scale dynamics in catalytic liquid systems and surface science.