Spectral Clustering to Analyze the Hidden Events in Single-Molecule Break Junctions

The single-molecule break junction technique provides a high-throughput method to explore the charge transport phenomena through a molecular junction at the ultimate scale of a single molecule. The most probable conductance of a molecular junction is normally extracted from histogram generated from repeated and massive break junction data. However, this conventional data analysis method only exhibits general charge transport properties of molecular junctions, and insightful information hidden in those recorded data remains unexplored. Among them, some of the conductance variations corresponding to different molecular junction conformations that occur during the break junction process might easily be overlooked. To accurately extract those hidden events, here we demonstrated a customized spectral clustering method with the evaluation of the Calinski–Harabasz index, which could be employed to analyze a large amount of data and to automatically extract different molecular junction conformations without subjective bias. Our approach was first validated through simulated data sets and was confirmed to be suitable for the product analysis during a chemical reaction. Moreover, using this method, an easily overlooked but unignorable junction conformation was found during the carborane molecular junction measurement, suggesting that spectral clustering with the Calinski–Harabasz index as a criterion offers a promising algorithm for junction conformation analysis in massive break junction data.