Automatic classification of single-molecule charge transport data with an unsupervised machine-learning algorithm

Abstract

Single-molecule electrical characterization reveals the events occurring at the nanoscale, which provides guidelines for molecular materials and devices. However, data analysis to extract valuable information from the nanoscale measurement data remained as a major challenge. Herein, an unsupervised deep leaning method, a deep auto-encoder K-means (DAK) algorithm, is developed to distinguish different events from single-molecule charge transport measurements. As validated by three single-molecule junction systems, the method applies to the recognition for multiple compounds with various events and offers an effective data analysis method to track reaction kinetics at the single-molecule scale. This work opens the possibility of using deep unsupervised approaches to studying the physical and chemical processes at the single-molecule level.