Theoretical Framework Based on Molecular Dynamics and Data Mining Analyses for the Study of Potential Energy Surfaces of Finite-Size Particles.

Abstract

Nanoclusters are remarkably promising for the capture and activation of small molecules for fuel production or as precursors for other chemicals of high commercial value. Since this process occurs under a wide variety of experimental conditions, an improved atomistic understanding of the stability and phase transitions of these systems will be key to the development of successful technological applications. In this work, we proposed a theoretical framework to explore the potential energy surface and configuration space of nanoclusters to map the most important morphologies presented by those systems and the phase transitions between them. A fully automated process was developed, which combines global optimization techniques, classical molecular dynamics, and unsupervised machine learning algorithms. To showcase these capabilities of the approach, we explored the example of copper nanoclusters (Cun) where n = 13, 38, 55, 75, 98, 102, and 147. We not only reported a graphical potential energy surface for each size, but also explored the topology of the configuration space via structural and thermodynamic analyses. The effect of size on the potential energy surface and the critical temperature for solid-liquid phase transitions were also reported, highlighting the impact of magic numbers on those quantities.