Grid-Based Correlation Analysis to Identify Rare Quantum Transport Behaviors

Abstract

Most single-molecule transport experiments produce large and stochastic datasets containing a wide range of behaviors, presenting both a challenge to their analysis, but also an opportunity for discovering new physical insights. Recently, several unsupervised clustering algorithms have been developed to help extract and separate distinct features from single-molecule transport data. However, these clustering approaches are in general neither designed nor appropriate for identifying very rare features and behaviors, such as switching events, chemical reactions, or particular binding modes, which may nonetheless contain physically meaningful information. In this work we introduce a completely new analysis framework specifically designed for rare event detection in singlemolecule break junction data as a necessary component to enable such studies in the future. Our approach leverages the concept of correlations of breaking traces with their own history to robustly identify paths through distanceconductance space that correspond to reproducible rare behaviors. As both a demonstrative and important example, we focus on rare conductance plateaus for short molecules, which can be essentially invisible when examining raw data. We show that our grid-based correlation tools successfully and reproducibly locate these rare plateaus in real experimental datasets. This result provides useful insight into the nanoscopic junction environment, enables a broader variety of molecules to be considered in the future, and validates our new approach as a powerful tool for detecting rare yet meaningful behaviors in single molecule transport data.

Beyond Simple Structure–Function Relationships: The Interplay of Geometry, Electronic Structure, and Molecule/Electrode Coupling in Single-Molecule Junctions

“…The experimental conductance of each OPV2 molecule from Scheme was measured using a custom-built mechanically controlled break junction (MCBJ) setup described previously. ,, Briefly, MCBJ samples were fabricated on a phosphor bronze substrate coated with an insulating layer of polyimide. A pattern containing a thin (∼100 nm) gold constriction was defined with electron beam lithography and then coated with 4 nm of titanium and 80 nm of gold.…”

Trusting our machines: validating machine learning models for single-molecule transport experiments

“…Several groups have already begun work on this and we think this is an important direction to explore. 15,74–76…”

“…Several groups have already begun work on this and we think this is an important direction to explore. 15,[74][75][76] A richer representation of the data would also enable more extensive use of traditional statistics. Perhaps an approach could be envisioned where newer machine learning techniques quickly filter out traces that has contaminants.…”

Making the Most of Nothing: One-Class Classification for Single-Molecule Transport Studies

“…Alternatively, methods have been developed that circumvent the need to remove blank traces entirely. For instance, Bamberger et al 37 have developed a method to extract traces with molecular features that were initially invisible in the combined raw data. They also developed a method to extract the main plateau of each trace, 38 which they later used to explain large variations in the reported conductance of different stilbene derivatives.…”