New Insights into Single-Molecule Junctions Using a Robust, Unsupervised Approach to Data Collection and Analysis

Methods for the analysis, interpretation, and prediction of single-molecule junction conductance behaviour

This article offers a broad overview of measurement methods in the field of molecular electronics, with a particular focus on the most common single-molecule junction fabrication techniques, the challenges in data analysis and interpretation of single-molecule junction current-distance traces, and a summary of simulations and predictive models aimed at establishing robust structure-property relationships of use in the further development of molecular electronics.

Machine-learning prediction of the formation of atomic gold wires by mechanically controlled break junctions

One of the challenging issues in the formation of atomic wires in break-junction experiments is the formation of stable monoatomic chains of reasonable length. To address this issue, in this study, we present a combination of unsupervised and supervised machine learning models trained on the experimentally measured conductance traces, with a goal to develop a microscopic understanding. Applying a machine learning model to two independent data sets from two different samples containing 72 000 and 90 000 conductance-displacement traces of single-atomic junctions, respectively, we first obtain the optimum conditions of bias and the stretching rate for the formation of chains of length > 4 Å. A deep learning method is subsequently applied for the classification of individual breaking traces, leading to the identification of trace features related to long-chain formation. Further investigation by ab initio molecular dynamics simulations provides a molecular-level understanding of the problem.

Measurements of single-molecule electromechanical properties based on atomic force microscopy fixed-junction technique

A hybrid technique combining atomic force microscopy and the fixed-junction technique is developed to simultaneously probe the electrical and mechanical characteristics of a single-molecule junction.

Plasmonic phenomena in molecular junctions: principles and applications

Molecular junctions are building blocks for constructing future nanoelectronic devices that enable the investigation of a broad range of electronic transport properties within nanoscale regions. Crossing both the nanoscopic and mesoscopic length scales, plasmonics lies at the intersection of the macroscopic photonics and nanoelectronics, owing to their capability of confining light to dimensions far below the diffraction limit. Research activities on plasmonic phenomena in molecular electronics started around 2010, and feedback between plasmons and molecular junctions has increased over the past years. These efforts can provide new insights into the near-field interaction and the corresponding tunability in properties, as well as resultant plasmon-based molecular devices. This Review presents the latest advancements of plasmonic resonances in molecular junctions and details the progress in plasmon excitation and plasmon coupling. We also highlight emerging experimental approaches to unravel the mechanisms behind the various types of light-matter interactions at molecular length scales, where quantum effects come into play. Finally, we discuss the potential of these plasmonic-electronic hybrid systems across various future applications, including sensing, photocatalysis, molecular trapping and active control of molecular switches.

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

In this tutorial review, we will describe crucial aspects related to the application of machine learning to help users avoid the most common pitfalls. The examples we present will be based on data from the field of molecular electronics, specifically single-molecule electron transport experiments, but the concepts and problems we explore will be sufficiently general for application in other fields with similar data. In the first part of the tutorial review, we will introduce the field of single-molecule transport, and provide an overview of the most common machine learning algorithms employed. In the second part of the tutorial review, we will show, through examples grounded in single-molecule transport, that the promises of machine learning can only be fulfilled by careful application. We will end the tutorial review with a discussion of where we, as a field, could go from here.

In this tutorial review, we will describe crucial aspects related to the application of machine learning to help users avoid the most common pitfalls.

Single-molecule junction spontaneously restored by DNA zipper

The electrical properties of DNA have been extensively investigated within the field of molecular electronics. Previous studies on this topic primarily focused on the transport phenomena in the static structure at thermodynamic equilibria. Consequently, the properties of higher-order structures of DNA and their structural changes associated with the design of single-molecule electronic devices have not been fully studied so far. This stems from the limitation that only extremely short DNA is available for electrical measurements, since the single-molecule conductance decreases sharply with the increase in the molecular length. Here, we report a DNA zipper configuration to form a single-molecule junction. The duplex is accommodated in a nanogap between metal electrodes in a configuration where the duplex is perpendicular to the nanogap axis. Electrical measurements reveal that the single-molecule junction of the 90-mer DNA zipper exhibits high conductance due to the delocalized π system. Moreover, we find an attractive self-restoring capability that the single-molecule junction can be repeatedly formed without full structural breakdown even after electrical failure. The DNA zipping strategy presented here provides a basis for novel designs of single-molecule junctions.

The versatility of DNA has inspired many single-molecule investigations utilizing nanotechnology. Harashima et al. have a somewhat different take on the subject and study a zipper configuration bridging electrodes that resembles an active electro-mechanical component instead.

Developing Longer-Lived Single Molecule Junctions with a Functional Flexible Electrode

Creating single-molecule junctions with a long-lived lifetime at room temperature is an open challenge. Finding simple and efficient approaches to increase the durability of single-molecule junction is also of practical value in molecular electronics. Here it is shown that a flexible gold-coated nanopipette electrode can be utilized in scanning tunneling microscope (STM) break-junction measurements to efficiently enhance the stability of molecular junctions by comparing with the measurements using conventional solid gold probes. The stabilizing effect of the flexible electrode displays anchor group dependence, which increases with the binding energy between the anchor group and gold. An empirical model is proposed and shows that the flexible electrode could promote stable binding geometries at the gold-molecule interface and slow down the junction breakage caused by the external perturbations, thereby extending the junction lifetime. Finally, it is demonstrated for the first time that the internal conduit of the flexible STM tip can be utilized for the controlled molecule delivery and molecular junction formation.

Single-Molecule Junctions with Highly Improved Stability

Single-molecule junctions (SMJs) have been fabricated using layers generated by diazonium electroreduction. This process creates a C-Au covalent bond between the molecule and the electrode. Rigid oligomers of variable length, based on porphyrin derivatives in their free base or cobalt complex forms, have been grafted on the surface. The conductance of the oligomers has been studied by a scanning tunneling microscopy break junction (STM-bj) technique and G(t) measurements, and the lifetime of the SMJs has been investigated. The conductance histograms indicate that charge transport in the porphyrins is relatively efficient and influenced by the presence of the cobalt center. With both systems, random telegraph G(t) signals are easily recorded, showing SMJ on/off states. The SMJs then stabilize and exhibit a surprisingly long lifetime around 10 s, and attenuation plots, obtained by both G(t) and STM-bj measurements, give identical values. This work shows that highly stable SMJs can be prepared using a diazonium grafting approach.

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

In order to design molecular electronic devices with high performance and stability, it is crucial to understand their structure-to-property relationships. Single-molecule break junction measurements yield a large number of conductance-distance traces, which are inherently highly stochastic. Here we propose a weakly supervised deep learning algorithm to classify and segment these conductance traces, a method that is mainly based on transfer learning with the pretrain-finetune technique. By exploiting the powerful feature extraction capabilities of the pretrained VGG-16 network, our convolutional neural network model not only achieves high accuracy in the classification of the conductance traces, but also segments precisely the conductance plateau from an entire trace with very few manually labeled traces. Thus, we can produce a more reliable estimation of the junction conductance and quantify the junction stability. These findings show that our model has achieved a better accuracy-to-manpower efficiency balance, opening up the possibility of using weakly supervised deep learning approaches in the studies of single-molecule junctions.

Achieving Ultralow, Zero, and Inverted Tunneling Attenuation Coefficients in Molecular Wires with Extended Conjugation

Molecular tunnel junctions are organic devices miniaturized to the molecular scale. They serve as a versatile toolbox that can systematically examine charge transport behaviors at the atomic level. The electrical conductance of the molecular wire that bridges the two electrodes in a junction is significantly influenced by its chemical structure, and an intrinsically poor conductance is a major barrier for practical applications toward integrating individual molecules into electronic circuitry. Therefore, highly conjugated molecular wires are attractive as active components for the next-generation electronic devices, owing to the narrow highest occupied molecular orbital-lowest occupied molecular orbital gaps provided by their extended π-building blocks. This article aims to highlight the significance of highly conductive molecular wires in molecular electronics, the structures of which are inspired from conductive organic polymers, and presents a body of discussion on molecular wires exhibiting ultralow, zero, or inverted attenuation of tunneling probability at different lengths, along with future directions.

Preparation and Application of Microelectrodes at the Single-Molecule Scale

Molecular electronics offers a potential solution for the miniaturization of electronics beyond conventional silicon electronics. A key goal of molecular electronics is to fabricate the single-molecule junction with the functions of electronic units. The term "molecular junction" means a molecular cluster or a single molecule incorporated between two microelectrodes, and electrons are transported across it. The methods of constructing molecular junctions dynamically were developed, such as STM-BJ, AFM-BJ, and MCBJ, providing precise control of the gap and easy measurement of thousands of junctions. Electrodes based on these techniques are commonly called microelectrodes because at least one dimension is on the micron scale. In this manuscript, we summarize the preparation methods of microelectrodes and their application in single-molecule measurements. In addition, we discuss the electrode factor that influences the molecular electrical properties, such as material, curvature radius and cone angle, and further provide a brief prospect of molecular electronics.

Long-Range Charge Transport in Diazonium-Based Single-Molecule Junctions

Thin layers of cobalt and ruthenium polypyridyl-oligomers with thicknesses between 2 and 8 nm were deposited on gold by electrochemical reduction of diazonium salts. A scanning tunneling microscope was used to create single-molecule junctions (SMJs). The charge transport properties of the Au-[Co(tpy)₂]_n-Au (n = 1-4) SMJs do not depend markedly on the oligomer length, have an extremely low attenuation factor (β ∼ 0.19 nm^-1), and do not show a thickness-dependent transition between two mechanisms. Resonant charge transport is proposed as the main transport mechanism. The SMJ conductance decreases by 1 order of magnitude upon changing the metal from Co to Ru. In Au-[Ru(tpy)₂]_n-Au and Au-[Ru(bpy)₃]_n-Au SMJs, a charge transport transition from direct tunneling to hopping is evidenced by a break in the length-dependent β-plot. The three different mechanisms observed are a clear molecular signature on transport in SMJs. Most importantly, these results are in good agreement with those obtained on large-area molecular junctions.

Using Deep Learning to Identify Molecular Junction Characteristics

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.

Unsupervised feature recognition in single-molecule break junction data

Single-molecule break junction measurements deliver a huge number of conductance vs. electrode separation traces. During such measurements, the target molecules may bind to the electrodes in different geometries, and the evolution and rupture of the single-molecule junction may also follow distinct trajectories. The unraveling of the various typical trace classes is a prerequisite to the proper physical interpretation of the data. Here we exploit the efficient feature recognition properties of neural networks to automatically find the relevant trace classes. To eliminate the need for manually labeled training data we apply a combined method, which automatically selects training traces according to the extreme values of principal component projections or some auxiliary measured quantities. Then the network captures the features of these characteristic traces and generalizes its inference to the entire dataset. The use of a simple neural network structure also enables a direct insight into the decision-making mechanism. We demonstrate that this combined machine learning method is efficient in the unsupervised recognition of unobvious, but highly relevant trace classes within low and room temperature gold-4,4' bipyridine-gold single-molecule break junction data.

Non-chemisorbed gold-sulfur binding prevails in self-assembled monolayers

Gold-thiol contacts are ubiquitous across the physical and biological sciences in connecting organic molecules to surfaces. When thiols bind to gold in self-assembled monolayers (SAMs) the fate of the hydrogen remains a subject of profound debate-with implications for our understanding of their physical properties, spectroscopic features and formation mechanism(s). Exploiting measurements of the transmission through a molecular junction, which is highly sensitive to the nature of the molecule-electrode contact, we demonstrate here that the nature of the gold-sulfur bond in SAMs can be probed via single-molecule conductance measurements. Critically, we find that SAM measurements of dithiol-terminated molecular junctions yield a significantly lower conductance than solution measurements of the same molecule. Through numerous control experiments, conductance noise analysis and transport calculations based on density functional theory, we show that the gold-sulfur bond in SAMs prepared from the solution deposition of dithiols does not have chemisorbed character, which strongly suggests that under these widely used preparation conditions the hydrogen is retained.

Cross-plane conductance through a graphene/molecular monolayer/Au sandwich

The functionalities offered by single-molecule electrical junctions are yet to be translated into monolayer or few-layer molecular films, where making effective and reproducible electrical contact is one of the challenging bottlenecks. Here we take a significant step in this direction by demonstrating that excellent electrical contact can be made with a monolayer biphenyl-4,4'-dithiol (BPDT) molecular film, sandwiched between gold and graphene electrodes. This sandwich device structure is advantageous, because the current flows through the molecules to the gold substrate in a 'cross-plane' manner, perpendicular to the plane of graphene, yielding high-conductance devices. We elucidate the nature of the cross-plane graphene/molecule/Au transport using quantum transport calculations and introduce a simple analytical model, which captures generic features of the current-voltage characteristic. Asymmetry in junction properties results from the disparity in electrode electrical properties, the alignment of the BPDT HOMO-LUMO energy levels and the specific characteristics of the graphene electrode. The experimental observation of scalability of junction properties within the junction area, in combination with a theoretical description of the transmission probability of the thiol-graphene contact, demonstrates that between 10% and 100% of the molecules make contact with the electrodes, which is several orders of magnitude greater than that achieved to date in the literature.

Fast and straightforward analysis approach of charge transport data in single molecule junctions

In this study, we introduce an efficient data sorting algorithm, including filters for noisy signals, conductance mapping for analyzing the most dominant conductance group and sub-population groups. The capacity of our data analysis process has also been corroborated on real experimental data sets of Au-1,6-hexanedithiol-Au and Au-1,8-octanedithiol-Au molecular junctions. The fully automated and unsupervised program requires less than one minute on a standard PC to sort the data and generate histograms. The resulting one-dimensional and two-dimensional log histograms give conductance values in good agreement with previous studies. Our algorithm is a straightforward, fast and user-friendly tool for single molecule charge transport data analysis. We also analyze the data in a form of a conductance map which can offer evidence for diversity in molecular conductance. The code for automatic data analysis is openly available, well-documented and ready to use, thereby offering a useful new tool for single molecule electronics.

Classification of conductance traces with recurrent neural networks

We present a new automated method for structural classification of the traces obtained in break junction experiments. Using recurrent neural networks trained on the traces of minimal cross-sectional area in molecular dynamics simulations, we successfully separate the traces into two classes: point contact or nanowire. This is done without any assumptions about the expected features of each class. The trained neural network is applied to experimental break junction conductance traces, and it separates the classes as well as the previously used experimental methods. The effect of using partial conductance traces is explored, and we show that the method performs equally well using full or partial traces (as long as the trace just prior to breaking is included). When only the initial part of the trace is included, the results are still better than random chance. Finally, we show that the neural network classification method can be used to classify experimental conductance traces without using simulated results for training, but instead training the network on a few representative experimental traces. This offers a tool to recognize some characteristic motifs of the traces, which can be hard to find by simple data selection algorithms.

Fast Data Sorting with Modified Principal Component Analysis to Distinguish Unique Single Molecular Break Junction Trajectories

A simple and fast analysis method to sort large data sets into groups with shared distinguishing characteristics is described and applied to single molecular break junction conductance versus electrode displacement data. The method, based on principal component analysis, successfully sorts data sets based on the projection of the data onto the first or second principal component of the correlation matrix without the need to assert any specific hypothesis about the expected features within the data. This is an improvement on the current correlation matrix analysis approach because it sorts data automatically, making it more objective and less time consuming, and our method is applicable to a wide range of multivariate data sets. Here the method is demonstrated on two systems. First, it is demonstrated on mixtures of two molecules with identical anchor groups and similar lengths, but either a π (high conductance) or a σ (low conductance) bridge. The mixed data are automatically sorted into two groups containing one molecule or the other. Second, it is demonstrated on break junction data measured with the π bridged molecule alone. Again, the method distinguishes between two groups. These groups are tentatively assigned to different geometries of the molecule in the junction.

Deep learning for single-molecule science

Exploring and making predictions based on single-molecule data can be challenging, not only due to the sheer size of the datasets, but also because a priori knowledge about the signal characteristics is typically limited and poor signal-to-noise ratio. For example, hypothesis-driven data exploration, informed by an expectation of the signal characteristics, can lead to interpretation bias or loss of information. Equally, even when the different data categories are known, e.g., the four bases in DNA sequencing, it is often difficult to know how to make best use of the available information content. The latest developments in machine learning (ML), so-called deep learning (DL) offer interesting, new avenues to address such challenges. In some applications, such as speech and image recognition, DL has been able to outperform conventional ML strategies and even human performance. However, to date DL has not been applied much in single-molecule science, presumably in part because relatively little is known about the 'internal workings' of such DL tools within single-molecule science as a field. In this Tutorial, we make an attempt to illustrate in a step-by-step guide how one of those, a convolutional neural network (CNN), may be used for base calling in DNA sequencing applications. We compare it with a SVM as a more conventional ML method, and discuss some of the strengths and weaknesses of the approach. In particular, a 'deep' neural network has many features of a 'black box', which has important implications on how we look at and interpret data.

Conformational Smear Characterization and Binning of Single-Molecule Conductance Measurements for Enhanced Molecular Recognition

Electronic conduction or charge transport through single molecules depends primarily on molecular structure and anchoring groups and forms the basis for a wide range of studies from molecular electronics to DNA sequencing. Several high-throughput nanoelectronic methods such as mechanical break junctions, nanopores, conductive atomic force microscopy, scanning tunneling break junctions, and static nanoscale electrodes are often used for measuring single-molecule conductance. In these measurements, "smearing" due to conformational changes and other entropic factors leads to large variances in the observed molecular conductance, especially in individual measurements. Here, we show a method for characterizing smear in single-molecule conductance measurements and demonstrate how binning measurements according to smear can significantly enhance the use of individual conductance measurements for molecular recognition. Using quantum point contact measurements on single nucleotides within DNA macromolecules, we demonstrate that the distance over which molecular junctions are maintained is a measure of smear, and the resulting variance in unbiased single measurements depends on this smear parameter. Our ability to identify individual DNA nucleotides at 20× coverage increases from 81.3% accuracy without smear analysis to 93.9% with smear characterization and binning (SCRIB). Furthermore, merely 7 conductance measurements (7× coverage) are needed to achieve 97.8% accuracy for DNA nucleotide recognition when only low molecular smear measurements are used, which represents a significant improvement over contemporary sequencing methods. These results have important implications in a broad range of molecular electronics applications from designing robust molecular switches to nanoelectronic DNA sequencing.

Reversible on-surface wiring of resistive circuits

Whilst most studies in single-molecule electronics involve components first synthesized ex situ, there is also great potential in exploiting chemical transformations to prepare devices in situ. Here, as a first step towards this goal, we conduct reversible reactions on monolayers to make and break covalent bonds between alkanes of different lengths, then measure the conductance of these molecules connected between electrodes using the scanning tunneling microscopy-based break junction (STM-BJ) method. In doing so, we develop the critical methodology required for assembling and disassembling surface-bound single-molecule circuits. We identify effective reaction conditions for surface-bound reagents, and importantly demonstrate that the electronic characteristics of wires created in situ agree with those created ex situ. Finally, we show that the STM-BJ technique is unique in its ability to definitively probe surface reaction yields both on a local (∼50 nm2) and pseudo-global (≥10 mm2) level. This investigation thus highlights a route to the construction and integration of more complex, and ultimately functional, surface-based single-molecule circuitry, as well as advancing a methodology that facilitates studies beyond the reach of traditional ex situ synthetic approaches.

Single-Molecule Transport at a Rectifying GaAs Contact

In most single- or few-molecule devices, the contact electrodes are simple ohmic resistors. Here we describe a new type of single-molecule device in which metal and semiconductor contact electrodes impart a function, namely, current rectification, which is then modified by a molecule bridging the gap. We study junctions with the structure Au STM tip/X/n-GaAs substrate, where "X" is either a simple alkanedithiol or a conjugated unit bearing thiol/methylthiol contacts, and we detect current jumps corresponding to the attachment and detachment of single molecules. From the magnitudes of the current jumps we can deduce values for the conductance decay constant with molecule length that agree well with values determined from Au/molecule/Au junctions. The ability to impart functionality to a single-molecule device through the properties of the contacts as well as through the properties of the molecule represents a significant extension of the single-molecule electronics "tool-box".

Single-molecule conductance of DNA gated and ungated by DNA-binding molecules

Single-molecule conductance can be controllably modulated by DNA-binding molecules.

Unsupervised vector-based classification of single-molecule charge transport data

The stochastic nature of single-molecule charge transport measurements requires collection of large data sets to capture the full complexity of a molecular system. Data analysis is then guided by certain expectations, for example, a plateau feature in the tunnelling current distance trace, and the molecular conductance extracted from suitable histogram analysis. However, differences in molecular conformation or electrode contact geometry, the number of molecules in the junction or dynamic effects may lead to very different molecular signatures. Since their manifestation is a priori unknown, an unsupervised classification algorithm, making no prior assumptions regarding the data is clearly desirable. Here we present such an approach based on multivariate pattern analysis and apply it to simulated and experimental single-molecule charge transport data. We demonstrate how different event shapes are clearly separated using this algorithm and how statistics about different event classes can be extracted, when conventional methods of analysis fail.

Conformations of cyclopentasilane stereoisomers control molecular junction conductance

Here we examine the impact of ring conformation on the charge transport characteristics of cyclic pentasilane structures bound to gold electrodes in single molecule junctions. We investigate the conductance properties of alkylated cyclopentasilane cis and trans stereoisomers substituted in the 1,3-position with methylthiomethyl electrode binding groups using both the scanning tunneling microscope-based break junction technique and density functional theory based ab initio calculations. In contrast with the linear ones, these cyclic silanes yield lower conductance values; calculations reveal that the constrained dihedral geometries occurring within the ring are suboptimal for σ-orbital delocalization, and therefore, conductance. Theoretical calculations reproduce the measured conductance trends for both cis and trans isomers and find several distinct conformations that are likely to form stable molecular junctions at room temperature. Due to the weakened σ-conjugation in the molecule, through-space interactions are found to contribute significantly to the conductance. This manuscript details the vast conformational flexibility in cyclopentasilanes and the tremendous impact it has on controlling conductance.