Using Deep Learning to Identify Molecular Junction Characteristics

Abstract

The scanning tunneling microscope-based break junction (STM-BJ) is used widely to create and characterize single metal-molecule-metal junctions. In this technique, conductance is continuously recorded as a metal point contact is broken in a solution of molecules. Conductance plateaus are seen when stable molecular junctions are formed. Typically, thousands of junctions are created and measured, yielding thousands of distinct conductance versus extension traces. However, such traces are rarely analyzed individually to recognize the types of junctions formed. Here, we present a deep learning-based method to identify molecular junctions and show that it performs better than several commonly used and recently reported techniques. We demonstrate molecular junction identification from mixed solution measurements with accuracies as high as 97%. We also apply this model to an in situ electric field-driven isomerization reaction of a [3]­cumulene to follow the reaction over time. Furthermore, we demonstrate that our model can remain accurate even when a key parameter, the average junction conductance, is eliminated from the analysis, showing that our model goes beyond conventional analysis in existing methods.

Deep Learning‐Enhanced Nanopore Sensing of Single‐Nanoparticle Translocation Dynamics

“…We now explain the deep learning approach (Figure 3d). It is based on a convolutional auto‐encoding neural network that operates repeated convolutions of input signals with each stage involving decision making on which features to employ in the next stage by using an activation function of either rectified linear unit (ReLU) or LeakyReLU (Figure S7, Supporting Information) [ 44,45 ] to virtually convert the I ion − t curves into low‐resolution yet feature‐enhanced datasets. Subsequently, the encoded input undergoes deconvolutions to reconstruct signals of original size.…”

Photoswitching Molecular Junctions: Platforms and Electrical Properties

“…Temperature dependence, anchoring group dependence (having an identical backbone), or electrode variation could provide more insights for mitigating the quenching effect [47] . To develop and optimize photochromic molecular species, machine learning can be leveraged, which has shown its potential for accelerated ab initio computation, reaction prediction, synthesis planning, and molecular junction identification [128–133] . Furthermore, new electrode materials have to be developed for higher optical transmittance and appropriate band structures including band alignment and contact coupling.…”

Using Weakly Supervised Deep Learning to Classify and Segment Single‐Molecule Break‐Junction Conductance Traces

“…Recently machine learning has been introduced into molecular electronics, [10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27] as a powerful tool to analyze break junction data. Supervised and unsupervised learning are the two main classes of machine learning algorithms, their main difference being whether or not a manually labeled training set is needed.…”

“…At the same time, supervised learning has also been applied to the classification of conductance traces. [23,24] For example, Lauritzen et al trained a recurrent neutral network for classifying experimental conductance curves of gold break junctions. [23] Moreover, a convolutional neural networks (CNN)-based method has been demonstrated to achieve a much higher accuracy in the identification of molecular junctions and a striking ability to sort conductance traces without relying on average conductance information.…”