Bridging the Gap: Using Machine Learning Force Fields to Simulate Gold Break Junctions at Pulling Speeds Closer to Experiments.

Abstract

The properties and dynamics of gold nanowires have been studied for decades as an important testbed for several physical phenomena. Gold nanowires forming at contacts are an integral part of molecular junctions used to study the electronic and thermal properties of single molecules. However, the huge discrepancy in time scales between experiments and simulations, compounded by the limited accuracy of classical force fields, has posed a challenge in accurately simulating realistic junctions. Here, we show that machine-learning force fields uncover phenomena not captured by classical force fields when modeling Au-Au pulling junctions. Our simulations show a dependency of the average breaking distance on the pulling speed, highlighting a more complex behavior than previously thought. Our results demonstrate that the use of more accurate force fields to simulate metallic nanowires is essential for capturing the complexity of their structural evolution in break junction experiments. Our developments advance the modeling accuracy of molecular junctions, bridging the gap between experimental and simulation time scales.