Clustering method of single-molecule electrical transport data based on spectral clustering

Abstract

The invention discloses a clustering method of single-molecule electrical transport data based on spectral clustering, and relates to the single-molecule electrical transport data. The method comprises the steps of collecting all single molecule conductance-distance curve data, and making a one-dimensional conductance columnar statistical graph for each piece of curve data; according to an interested conductance interval, selecting the conductance columnar statistical graph of the corresponding interval as a data set of spectral clustering; constructing a similarity matrix A according to an algorithm flow of spectral clustering; selecting a clustering number K from 2 to M by utilizing a standard flow of spectral clustering, and clustering the one-dimensional conductivity columnar statistical graph into K classes; obtaining an optimal clustering number Kopt according to the maximum value of the CH index; and reserving a clustering result with the clustering number of Kopt to obtain a final optimal clustering result of the monomolecular conductance-distance curve data. In the process of measuring the conductivity of a certain single molecule, several sets of most possible conductivity values of the single molecule can be accurately and effectively obtained, and the reliable data information is provided for researching the electrical transport property of the single molecule.

Description

technical field [0001] The invention relates to single-molecule electrical transport data, in particular to a clustering method for single-molecule electrical transport data based on spectral clustering. Background technique [0002] Currently, Scanning Tunneling Microscope Break Junction (STM-BJ) technology and Mechanically Controllable Break Junction (MCBJ) technology are common techniques for measuring single-molecule conductance. Electrode pairs can repeatedly acquire a large amount of conductance-distance curve data in a single experiment, that is, single-molecule electrical transport data. For such a large amount of conductance data, efficient and credible data processing and characterization will help to fully mine the information related to single-molecule conductance in the data, and provide data basis for the preparation of molecular devices. [0003] The traditional one-dimensional conductance histogram and the two-dimensional conductance-distance histogram still...

AI Technical Summary

Problems solved by technology

[0003] The traditional one-dimensional conductance histogram and the two-dimensional conductance-distance histogram still play an important role, but this statistical method based on all data inevitably has its inherent problems: 1. Small events It is easy to be covered by a large trend, resulting in failure to be noticed; 2. When multiple events occur with the same probability, this method is prone to produce wrong information; 3. Due to the variability of molecular junction conductance, this method cannot give How many specific events can occur in the knot experiment

In 2016, Mario Lemmer et al. (Lemmer, M.; Inkpen, M.S.; Kornysheva, K.; Long, N.J.; Albrecht, T. Nature Communications 2016, 7) proposed a vector-based multi-parameter classification method (Multi-Parameters Vector -based Classification process, MPVC) groups different electrical signals in single-molecule electrical transport data, but has poor screening results for data with multiple sets of conductance steps

In 2018, Hamill et al. (Hamill, J.M.; Zhao, X.T.; Meszaros, G.; Bryce, M.R.; Arenz, M. Phys Rev Lett 2018, 120, 016601) proposed a principal component analysis method (Principal Component Analysis, PCA ) to achieve the screening of two types of conductance step data, which is limited to two types and has certain limitations

However, none of these methods have specific and effective evaluation indicators, and they all obtain the number of possible clusters in the data set based on experience.

Images