Charge transport through molecular ensembles: Recent progress in molecular electronics

This review focuses on molecular ensemble junctions in which the individual molecules of a monolayer each span two electrodes. This geometry favors quantum mechanical tunneling as the dominant mechanism of charge transport, which translates perturbances on the scale of bond lengths into nonlinear electrical responses. The ability to affect these responses at low voltages and with a variety of inputs, such as de/protonation, photon absorption, isomerization, oxidation/reduction, etc., creates the possibility to fabricate molecule-scale electronic devices that augment; extend; and, in some cases, outperform conventional semiconductor-based electronics. Moreover, these molecular devices, in part, fabricate themselves by deﬁning single-nanometer features with atomic precision via self-assembly. Although these junctions share many properties with single-molecule junctions, they also possess unique properties that present a different set of problems and exhibit unique properties. The primary trade-off of ensemble junctions is complexity for functionality; disordered molecular ensembles are signiﬁ-cantly more difﬁcult to model, particularly atomistically, but they are static and can be incorporated into integrated circuits. Progress toward useful functionality has accelerated in recent years, concomitant with deeper scientiﬁc insight into the mediation of charge transport by ensembles of molecules and experimental platforms that enable empirical studies to control for defects and artifacts. This review separates junctions by the trade-offs, complexity, and sensitivity of their constituents; the bottom electrode to which the ensembles are anchored and the nature of the anchoring chemistry both chemically and with respect to electronic coupling; the molecular layer and the relationship among electronic structure, mechanism of charge transport


I. INTRODUCTION
Molecular electronics (ME) describes the field of research in which single molecules or ensembles of molecules are utilized as functional elements in electronic circuits.In 1971, Mann and Kuhn 1 successfully measured conductance through monolayers of cadmium salts of fatty acids.Though not the first to measure the conductance of molecular ensembles, 2 they observed a temperature-independent, exponential decrease in conductivity with increasing thickness of the monolayers, which they ascribed to incoherent tunneling through the organic monolayers.However, the modern concept of ME was introduced in 1974 when Aviram and Ratner 3 published a theoretical paper describing tunneling charge transport through a single molecule comprising donor and acceptor moieties that, when connected to two electrodes, acted as a molecular rectifier.Their work provided an ad hoc approach to predict the charge transport through molecules, which set the stage for subsequent studies that led to such modern approaches as those combining density functional theory (DFT) with the nonequilibrium Green's function (NEGF) approach to ME. [4][5][6] In those early years, the primary experimental hurdle was the attachment of electrodes to individual molecules; while Mann and Kuhn used Hg to contact Langmuir-Blodgett films, it was the invention and development of scanning tunneling microscopy (STM) and atomic force microscopy (AFM) in the 1980s that laid the foundation for modern research on molecular ensemble junctions (MEJs) by allowing the interrogation of individual molecules (but not MEJs themselves) on surfaces.0][11][12] These pioneering studies advanced the methodology for determining the electrical conductivity of different molecules and provided insight into their charge-transport properties in experimental platforms that were readily reproducible in different laboratories.These early experiments drove broad interest in cross-disciplinary physical scientists, which has led to the acceleration of fruitful discoveries of recent decades.
This review focuses specifically on the role of MEJs in ME because they are central to both the theoretical advancements and the development of useful devices that are reflected in the two complementary goals of ME: understanding the underlying physical properties and extracting useful functionality, i.e., constructing devices.8][19] The principal goal of the former is to develop a fundamental understanding of electron transport through junctions comprising single molecules, a construct that is relatively straightforward to treat in silico.The principal goal of the latter is functionality on the device level, utilizing ensembles of molecules, usually in the form of self-assembled monolayers (SAMs), to define both the function and the smallest dimension of the junction.Singlemolecule techniques like break junctions (BJs), in which the electric leads repeatedly form and rupture by the trapping and releasing of individual molecules, are primarily useful as spectroscopic tools for investigating the detailed mechanisms of charge transport; they are not (yet) useful as devices because they are transient, but they can be modeled atomistically.This review focuses on (large-area) junctions (as summarized in Fig. 1) comprising molecular ensembles, which are potentially directly useful in technological applications, as they are intrinsically static, solid-state devices; however, they are difficult to model.
0][21] Our goal is to review the recent progress in this field and give the reader a general idea of what has been done, with an emphasis on the types and properties of molecular junctions.In Sec.I, we briefly discuss the mechanisms of charge transport, A conceptual summary of molecular ensemble junctions.An ensemble of molecules, often a single monolayer, is sandwiched between two electrodes (gray and gold spheres) to form a junction that can be completely symmetric, utilize electrodes of different materials, and/or comprise asymmetric molecules.The constituent molecules can be alkanes, shown here with two common types of anchoring groups, p-conjugated molecules, organometallics, or biomolecules.In each case, the molecular properties translate into useful electrical outputs, such as switching and rectification.
followed by a discussion of issues that are specific to MEJs.We also touch on the interpretation of data and the statistics used to analyze primary data with the aim of clarifying differences intrinsic to types of junctions.The review is presented from the perspective of forming a molecular junction: the bottom electrode and bottom interfaces are reviewed in Sec.II, molecular monolayers are reviewed in Sec.III, and top electrode and top interfaces are reviewed in Sec.IV.Finally, we briefly discuss the future outlook of ME in Sec.V.

A. Mechanisms of charge transport
The mechanisms of charge transport through ensembles/SAMs can be broadly divided into two phenomenological extremes: tunneling and hopping.The major differentiator is whether the chargetransport process is thermally activated.Charge transport through a molecular junction can be described as the propagation of an electron wave between electrodes that is modulated by a molecule/molecular layer between them.When the energy and phase of the charge carrier are maintained during the process, the mechanism is considered to be tunneling.At the other extreme is hopping, in which the charge relaxes inside the junction, requiring thermal activation for traversal.In general, tunneling preponderates across shorter distances (i.e., molecules) and hopping across longer distances. 22,23][29] The most commonly used theoretical model to describe temperature-independent tunneling is the simplified Simmons model, [30][31][32] which considers the junction formed by the molecule as an approximately trapezoidal potential barrier and is mathematically described by Eq. 1 where J is the current density; V is the applied bias; d is the zero-bias width of the barrier, which is nominally determined by molecular length; b is the tunneling decay coefficient and scales with the square root of the difference between the Fermi level E f and the frontier orbitals of the molecule(s); and J 0 is the theoretical value of J when d ¼ 0. This model captures the length dependence of J in relatively simple systems, such as SAMs of alkanethiols, [33][34][35] but its most important feature is that the values of b are highly reproducible across laboratories and experimental platforms.While the Simmons model is too simple to provide much meaningful physical insight, b and (to a lesser extent) J 0 are invaluable benchmarks for validating experimental platforms and controlling for their specific effects. 17,36It presupposes tunneling; for thermally activated charge transport, the relationship between current density and temperature is expressed as in Eq. 2, where E a is the activation energy of the hopping step.In practice, b is much more commonly reported because it is easy to obtain from any platform in which at least three molecules of different lengths can be measured, and it does not require variable temperature measurements.

B. Characteristics of molecular ensemble junctions
There are many phenomena that are specific to MEJs (compared to SMJs), for example, intermolecular effects [37][38][39] and the impositions of molecular self-assembly. 40Molecular ensemble junctions are also very sensitive to the topography of the metal substrate (i.e., the bottom electrode), and defects in molecular monolayers cannot be avoided due to the ubiquity of grain boundaries, step edges, and impurities in the substrate. 20Weiss et al. 34 fabricated junctions comprising SAMs of alkanethiolates on rough and smooth substrates using a hanging Hg drop top electrode, observing that thin-area defects and thick-area defects cause the overall conductance to be higher and lower, respectively, than nondefect junctions, readily revealing defects by direct comparisons to benchmark MEJs.These defects are defined by a change in d (from Eq. 1) caused by substrate topography and/or the morphology of the SAM vis-a-vis deviations from ideal packing.Jiang et al. 41 further observed that these defects can scale with the junction area, which can cause Joule heating in the junctions that affect J 0 rather than J. Thus, J/V curves alone cannot necessarily reveal defects present in an MEJ because changes to J 0 caused by defects that are systematic or endemic to a particular experimental platform affect the accuracy of J by shifting the J/V curves uniformly.(Accuracy characterizes the distance between a measurement and its true value.)Ordered molecular ensembles like SAMs also exhibit collective effects that are independent of the substrate and that cannot be observed in SMJs, for example, odd-even effects driven by the conformation of close-packed alkyl chains, 35,[42][43][44][45] the collective action of molecular dipoles affecting transport by shifting vacuum levels, [46][47][48][49][50] and quantum interference (QI) effects (which are described in Sec.III) correlated with molecular conformations imposed by self-assembled monolayers. 51Although many intermolecular effects are system specific, Dubi 52 showed that incorporating in-plane dephasing into transport calculations helps to address several universal transport features of ensemble junctions, such as the exponential decay of current with molecular length, the odd-even effect, and negative differential resistance.Thus, some of the phenomena that are specific to MEJs are the result of practical constraints, not physical differences.

C. Data collection and analysis
Molecular ensemble junctions are typically measured with one electrode grounded while current is recorded as a function of applied bias to produce a J/V trace.However, due to the aforementioned phenomena, J/V traces acquired from different areas on a substrate or across multiple substrates can vary by several orders of magnitude.Treating this variance correctly is integral to extracting physical insight from J/V data, and reducing it is vital to potential technological applications.Moreover, tunneling currents through SAMs are often in the pA regime at low bias and can increase by many orders of magnitude over bias windows of only 1-2 V. Combined, these features of MEJs can lead to noisy data, which is obviously an impediment to both research and technological application.In some cases, additional noise is intrinsic to the type of measurement 17 because of poor experimental design or inadequate control of experimental variables.Without appropriate and rigorous standards for measurement, incorrect conclusions can be drawn from improper data selection or winnowing from large numbers of failed junctions.For example, using the liquid metal eutectic Ga-In (EGaIn), the experience of the operators demonstrably impacted the precision-in practical terms, the size of the error bars-of the measurements; 35,53 a less experienced operator is more likely to encounter a lower yield of non-shorting junctions and broader spreads in J across junctions and substrates.While the accuracy of J is not user dependent, subtle effects can be masked by low precision.Of greater concern, defects in the substrate and the environment under which a junction is measured can have pronounced effects, 54 even concealing/revealing length dependence entirely. 55sing SAM-based junctions with EGaIn electrodes as an example, data are typically collected across multiple substrates.For each substrate, at least n % 10 cycles are measured from 0 V !V max ðþÞ !V max ðÀÞ !0 V such that at least 2n traces/retraces are measured for each of N % 10 junctions.(Here, a junction is equivalent to a single contact formed between a tip and a substrate.)Junctions that do not short over all the n cycles are considered non-shorting for computing yields.Averaged J/V curves are constructed from Gaussian fits to the log-normal distributions of J for each value of V as depicted in Fig. 2.This method of data collection assumes that the variance in J is the result of randomly distributed defects that exponentially affect J by locally varying d from Eq. 1. 34 That is why it is important that the N junctions should be measured at different areas of each of several substrates that are preferably prepared on different days, use different monolayer-forming solutions, etc.The details can vary across or within laboratories but are necessarily standardized in order to produce meaningful statistics and are reliable as long as the data within a particular study were collected and analyzed using the same procedure.Simeone et al. 36 observed that data collected with a nonstandardized protocol-forming an arbitrary number of junctions, running arbitrary numbers of scans-can decrease precision by increasing the variance in J.They proposed a "1/3/20 protocol" for the collection of data: For each freshly formed EGaIn electrode, 56,57 they formed three junctions in three different places on a Ag substrate supporting a SAM; for each junction, they recorded 20 J/V traces/retraces.They claim that this 1/3/20 protocol can avoid the effects of adventitious impurities on the EGaIn electrode, and collecting the same number of J/V curves per EGaIn electrode ensures that each freshly prepared electrode has the same statistical weight.The 20 scans are necessary so that one does not statistically over-weight instabilities of the current that can occur in the first few scans, as the electrode conforms to the topography of the SAM. 58Although the specifics vary with the experimental platform, for example, an Au-coated AFM tip does not oxidize in air, the statistical methods are universally applicable for ensuring precise, accurate measurements.
The choice of statistical methods used, which analyzes the large, often noisy datasets produced by MEJs, directly impacts the interpretation of the data.Since the source of error is the defects (as defined two paragraphs previously) that are distributed randomly, population statistics can be used to distinguish between distributions of J that differ by coincidence but belong to the same parent population (i.e., measurement artifacts) and those that differ because they are representative of different parent populations (i.e., that are caused by physical differences between MEJs).Reus et al. 58 examined different statistical methods, summarized in Fig. 3, that are applicable to such data, specifically those produced by measurements of SAMs with EGaIn electrodes.Methods 1-3 use the data to calculate single-compound statistics: Method 1 fits Gaussian functions to histograms of log jJj, method 2 uses the median and interquartile range, and method 3 uses arithmetic means and standard deviations (SDs).Methods 4a and 4b proceed directly to plotting and fitting the raw data to determine trend statistics; method 4a uses an algorithm that minimizes the sum of the absolute values of the errors between the data and a fit (a least absolute errors algorithm), while method 4b employs an algorithm that minimizes the sum of the squares of those errors (a least squares algorithm).They compared the accuracy and precision of these methods for individual SAMs, between two or more SAMs, and to determine b and J 0 .They concluded that methods 1, 2, and 4a are all acceptably accurate, while methods 3 and 4b are not.Method 1 assumes informative measurements of log jJj, that Eq. 1 is a valid, that the data are distributed log-normally, and that any deviations of log jJj from normality are not informative, for example, the data marked "short" in Fig. 2. Methods 2 and 4a assume that informative measurements of log jJj represent the bulk of the data and that noninformative data comprise extreme values.They favor method 1 over method 2, but both are probably accurate enough to use in reporting singlecompound statistics.Method 4a is as accurate as method 2 and about an order of magnitude more precise.They recommend method 4a for calculating trend statistics as long as the result does not conflict with method 1 and method 2. They do not recommend methods 3 and 4b, as those methods respond too strongly to noninformative, extreme values of log jJj.
In practice, most MEJ data are reported using method 1 because it is straightforward and because most MEJ data fit the criteria for its use.Although method 1 mathematically excludes extreme values of J, it also remains common practice to preselect data because of the aesthetics of digital publishing vis-a-vis the graphical representation of data.A more statistically sound approach that accomplishes the same goal is to plot the error as the confidence interval (CI) instead of the SD using Eq. 3, where z is no standardized the inverse of the cumulative distribution function for the standard normal distribution-the ttest parameter-and is based on the confidence level (usually 95% or 99%) and N is the number of junctions (not J/V traces) measured. 58ata are then plotted as the Gaussian mean value of the current density hJi6CI instead of hJi6SD.This method uses the Student's t test because the true standard deviation of the population is not known.
The practical consequence of this methodology is that one cannot say that two values of hJi whose CI overlaps come from two different populations, which is not the case for SD.In other words, the true value of hJi is likely to reside within the CI, so overlapping error bars computed from CIs imply that hJi can differ by coincidence rather than by an underlying physical difference.It is also important to distinguish between values of hJi reported with SD, CI, and standard error (of the mean).When the goal is to describe the underlying statistical distribution, as is the case with method 1, the error should be reported as SD.
When reported as CI, the error reflects an estimation of the range of hJi if the measurement were repeated in the future.Both reflect the precision of the measurement, which is the definition of error bars that is typically assumed in the MEJ literature.Standard error, however, reflects the accuracy of a measurement and is more common in SMJs, where the assumption is that the mean conductance reflects the most probable configuration of the junction, which has only one true value.Statistical methods are, unfortunately, not always reported with sufficient detail to ascertain which method is being used and how error is computed, which complicates meta-analyses.
Sporrer et al. 59 reported the use of heat maps of raw data to expose information that is otherwise concealed by statistical treatments.They intentionally prepared SAMs under experimental conditions that would affect their quality and collected data via users with different levels of expertise.The raw data, without preselection, were plotted in both 3D and 2D formats, as shown in Fig. 4.Although informative, 3D plots can be difficult to interpret, while a 2D heat map-a planar version of the 3D data-captures the skewness and outliers in a single figure.(Skewness is indicative of a shift in the mode and median values of a population relative to its mean.)The two dominant distributions, as in the bottom panel in Fig. 4(a), suggests a skewness, which the authors hypothesized as an indication of the origin of the adjacent major distribution, with the higher log jJj values from thin-area defects and the lower log jJj values from thick-area defects.They further validated this assumption by looking at the data obtained from junctions using soft/hard contact methods that are known to induce thick/thinarea defects.These data indeed showed skewness toward low/high values of current densities commensurate with the type of defect.The same authors later reported that the analysis of experimental datasets with higher statistical moments (skewness and kurtosis) reveals the dynamic nature of the MEJ. 60(Kurtosis describes the convergence toward the mean, the "peakedness," of a population.)They investigated MEJs comprising SAMs of molecules with two dipole moments, an internal amide and varied terminal R groups in which intramolecular Keesom (dipole-dipole) interactions were revealed by analysis of skewness and kurtosis; molecules bearing more-polarizable moieties exhibited an applied field-induced negative skewness in log jJj with concomitant narrowing of the tails (larger kurtotic value), while moieties bearing fixed dipole moments did not show these correlations.
While the aforementioned methods deal with primary J/V data, Vilan 61 used normalized differential conductance (NDC) to explore the effects of applied bias on the detailed mechanism of tunneling.The author showed the power of NDC plots in revealing subtle features that are associated with specific tunneling models.Mathematical modeling of the tunneling process will differ depending on the assumptions made in the analysis.Two well-known examples are the single-level model [62][63][64][65] and the (aforementioned) Simmons model.These models have very different mathematical expressions, although they yield I/V (or J/V) traces that are graphically very similar. 66owever, they exhibit subtle characteristic features that are difficult to observe on a linear scale but can be revealed in NDC plots.
Additionally, NDC analysis can be used to extract quantitative information, such as scaling bias V 0 , which is sensitive to the choice of the tunneling model.The author further showed the practical utility of NDC by analyzing the experimental data from MEJs comprising SAMs of short alkanethiols on smooth and rough Ag substrates, though NDC analysis can be used for any MEJ and is valid for tunneling charge transport in general.
The statistical techniques described herein are valid for any experimental technique that measures molecular ensembles.While some of the specific methods are specific to EGaIn (e.g., how often a tip should be regenerated), measurements on ensembles will always lead to some variance.That variance can be very narrow, which in fact is a prerequisite of technological applications.Under-sampling, due to low yields of junctions or high fabrication overhead, can adversely affect accuracy and precision, for example, increasing the latter (small spreads in data) and decreasing the former (large junction-to-junction or device-to-device variation).In all cases, the proper application of rigorous statistics is necessary to ensure that the true population is described by the data, which is vital for reproducibility, validation of theoretical models, and technological applications.

II. THE BOTTOM ELECTRODE
Molecular ensemble junctions comprise three distinct regions that combine to define the overall electrical properties and that can be manipulated independently of each other: (i) the interface between the bottom electrode (e.g., the substrate that supports a SAM) and the anchoring group that affixes the molecules to the electrode, (ii) the backbones of the molecules, and (iii) the interface between the top electrode and the terminal group of the molecule [which is identical to (i) in a symmetrical junction].While SMJs can be gated by adding a third electrode, 67 ensemble junctions must be gated from above/ below 68 or beside/in addition to 69 the molecular layer to affect the field uniformly.External stimuli like electrostatic, optical, or chemical gates can operate on (i), (ii), or (iii).The modulation (i.e., by external stimuli, synthesis/fabrication, or self-assembly) and combination of the three components give rise to the phenomena-many of which are unique to MEJs-that underpin (potential) applications of ME.We discuss each of these components separately, addressing fundamental properties, recent advances and challenges, and future potential.
Analogous to cell membranes, which mediate communications between the interior and exterior of the cells, preserve their physical integrity, and give them shape, 70 interfaces in MEJs mediate the flow of charge in and out of the junction, maintain the electric field across the molecular layer, and define the dimensions of the junction.The understanding of the bottom interface in molecular junctions has steadily improved over time.2][73][74] In this section, we survey the development of substrates and binding motifs in MEJs and examine the role of the bottom interface in practical molecular electronic devices.

A. Thiols and their derivatives
Due to their relative simplicity, their reproducibility, and the versatility of organic and organometallic compounds, SAMs of thiols have found applications across many scientific and engineering disciplines, e.g., anti-fouling surfaces, 75 lubrication, 76,77 corrosion resistance, 78,79 protein binding, 80,81 DNA assemblies, 82 cellular signaling and interactions, 83 photovoltaics, [84][85][86] and transistors. 68,87In MEJs, thiols (and/or disulfides) serve as anchoring groups to form molecular  60 ensembles, which then define the smallest dimensions of electrical junctions either as self-standing SAMs 7,88 or in the direction of the sequential self-assembly of other molecules (particularly biomolecules).This is a key feature of SAMs in particular because they spontaneously form supramolecular structures with very large aspect ratios, define the distance between the top and bottom electrodes, and intrinsically couple the macro world to the molecular world without the need for complex subtractive manufacturing to predefine the electrodes.Thiols are particularly important in this regard due to their tolerance toward various conditions.They will spontaneously form SAMs from the gas phase or from solution through the reversible, but strong interactions with various noble metals and transition metals, including Au, [88][89][90][91][92][93][94][95] Ag, 24,35,36,[96][97][98][99] Cu, [100][101][102] Hg, 103,104 Pd, [105][106][107] Pt, 74,108,109 Ni, [110][111][112][113] and Fe. 79,114Although the surface attachment chemistry can vary, thiol-SAMs are not limited to (coinage) metals and will form on semiconductor materials, such as GaAs, [115][116][117] InP, 118,119 MoS 2 , and WSe 2 . 120Other sulfur-containing groups, such as sulfides, disulfides, [121][122][123][124] and thiocyanates, 125,126 are also capable of forming SAMs.However, despite their versatility over the selection of substrates, thiols and their derivatives do not form SAMs on the surfaces of oxides.
The ubiquity of thiol-SAMs and their derivatives in other fields of research is beneficial to their use in MEJs because methods for investigating their properties have been extensively studied.Over the past three decades, the kinetics of SAM formation, 127 the details of their structure and defects, 20,89,[128][129][130][131] their charge-transport properties, 19 and their applications 21 have been thoroughly studied.Simple SAMs of alkanethiols are particularly well understood.They form through a multi-step process, and the monolayer is 90% complete (e.g., with respect to the final thermodynamic structure) within the first minute of growth in a fast first step and is only able to reach equilibrium after several hours in a slower second step.Thiols functionalized with a longer backbone (e.g., n-alkanethiols C n SH, n > 8) generally favor the formation of more ordered SAMs than shorter ones, though this is largely a function of their melting points, and SAMs of short alkanethiols are, in some respects, more robust than their longer counterparts. 423][134] During the formation of the SAM, small islands of thiol-functionalized molecules are first generated that merge into larger, continuous regions in a "striped phase" that can clean and reorganize metallic surfaces as they become more densely packed and eventually form the final, standing-up phase.The quality of SAMs formed from disulfides is comparable to those using thiols, despite slightly slower kinetics (% 40%) and the cleavage of S-S bond after the initial physisorption of the molecules. 135In contrast, the adsorption rate of sulfides is three orders of magnitude smaller than those of thiols and disulfides, and the SAMs of sulfides exhibit a lower coverage with a substantial amount of contamination on the surface of the substrates.In the case of thiocyanate-anchoring groups, the cleavage of S-CN bonds results in the assembly of molecules and the generation of adsorbed metal-cyanide species that are incorporated into the monolayer and can prevent full coverage of the sequentially formed thiolates. 125Similarly, symmetric, conjugated molecules functionalized by thioacetates result in poorly packed low-coverage SAMs by the partial deprotection of anchoring groups at the substrate; in contrast, the choice of bases provides a better control over the deprotection of thioacetates and facilitates the formation of dense, high-quality SAMs. 40,51ecent studies by Jiang et al. 136 on the SAMs of disulfides or thioacetates functionalized with bulky head groups (e.g., ferrocenes) revealed a drastically different picture of their structural integrity.As determined by wet chemistry and x-ray photoelectron spectroscopy (XPS), these SAMs are 2-10 times lower in surface coverage compared to SAMs of thiols and have a higher density of defects; however, they do not affect the yield of non-shorting junctions but rather the loss of other useful electrical properties, including rectification as a result of asymmetric charge transport through the SAMs.

Self-assembly
The rearrangement of the metal-molecule interface by thiols and their derivatives during the self-assembly process on metals plays a critical role in the formation of the SAMs as well as their properties in MEJs.Thiolate species occupy either fourfold (e.g., on Cu) or threefold (e.g., on Ag and Au) coordinated hollow sites and form metal-thiolate adlayer moieties that are less densely packed than the bulk metal. 137,138his reconstruction of the outermost atomic layer is often accompanied by intermolecular interactions that facilitate the formation of ordered SAMs and is capable of exchanging with free thiols from solution to repair defects; the metal-thiolate adlayer is able to minimize its free energy by reorganizing the metal surface and the packing of the molecular backbones.0][141] Mixed monolayers can be formed either by exposing substrates to mixtures of thiols or by exchanging thiols into existing SAMs, so-called in-place exchange.For example, bulky thiols bearing moieties that impart electrical function to an MEJ cannot form ordered SAMs on their own because the driving force to passivate the surface is too strong to leave the vacancies that would be required to accommodate the packing of the head groups. 20Bulky thiols can, however, be incorporated into mixed monolayers via in-place exchange such that the surface is passivated by homogenous domains of short alkanethiols perforated by bulky groups. 25,38,136Figure 5 shows a mixed monolayer that is able to assimilate fullerene cages, which are $0.7 nm in diameter.The appearance of rectification is both dependent on and a proof of the mixed monolayer.This same strategy has been employed on spiropyran photoswitches that form over-crowded SAMs with low surface coverage. 142The resulting steric congestion inhibits the ring-opening of the spiropyran moieties, dampening the electrical response resulting from the difference in conductance between the spiropyran and ringopened merocyanine forms.The formation of a mixed monolayer of the spiropyran mitigates the steric hindrance, maximizing the on-off ratio in the MEJ and reducing the fatiguing of the switches.These mixed monolayers were formed in reverse order compared to the fullerene, with SAMs of pure spiropyran being incubated in solutions of short alkanethiols.A detailed study of this process by Kumar et al. 124 using XPS revealed that the alkanethiols undergo metathesis with the weakly bound disulfides that anchor the spiropyrans, converting them to metal-thiolate bonds as they reorganize the SAM into a mixed monolayer.This shows that in-place exchange is a generalizable strategy for producing high-quality SAMs by removing disulfides as well as for forming mixed monolayers.In-place exchange can be an option for forming SAMs from molecules bearing multiple functional groups that can compete with thiols; pre-passivating a substrate with alkanethiols and then exchanging the desired thiol precludes any surface interaction other than the exchange of thiols.

Thiol-metal bonding
Apart from the formation of molecular ensembles, the interface between the bottom electrode and the anchoring groups (e.g., thiols and their derivatives) also has a significant impact on the chargetransport properties of the MEJ.4][145][146] Together with the embedded dipole in the molecular tails, they enable the modification of the Fermi level E f (which is equivalent to and sometimes reported as a shift in work function or vacuum level) of the bottom electrode by rational synthesis, reshaping the coupling between the energy levels of molecules and electrodes and giving rise to interesting phenomena in MEJs.In SAMs of oligophenylene thiols and oligoacene thiols, for example, the offset between the frontier molecular orbitals and E f (which is correlated with the height of the tunneling barrier) decreases, and the strength of coupling between the molecule and the electrode (which contributes to the broadening of the energy levels of frontier orbitals) increases, resulting in an increasing electrical conductance in MEJs comprising Pt, Au, and Ag electrodes. 109,145The nature of the interaction between the molecules and the electrode significantly impacts the charge-transport properties of MEJs.An interface formed by covalent bonds (chemisorption) produces stronger coupling, which tends to cause higher electrical conductance than those formed by van der Waals interactions (physisorption).Interestingly, although it is broadly acknowledged that the formation of thiol-SAMs is driven by the chemisorption of molecules, a recent study by Inkpen et al. 147 on the charge transport through thiol-SAMs by STM-BJs, in which molecules are repeatedly trapped in and released from SMJs formed between the STM tip and the bottom electrode, claimed that thiol-SAMs do not predominantly comprise chemisorbed molecules.The emergence of different singlemolecule conductance between thiol-SAMs measured in solution and dry conditions in this study is explained by the presence of excess physisorbed thiols in SAMs from the latter, which results in the preservation of hydrogen in the thiol-anchoring group. 148These results agree with subsequent claims that so-called "unbound sulfur species" that are identifiable by a characteristic XPS peak are, in fact, non-covalently bound thiol-metal complexes that are converted to covalent metalthiolates through metathesis with free thiols from solution. 124These non-covalent and covalent forms of thiol-SAMs are also distinguishable in MEJs by their length dependence (b).It may be difficult to reach a consensus on the true nature of binding chemistry in thiol-SAMs due to the unavoidable over-interpretation of data from disparate types of junctions and spectroscopes (e.g., STM-BJ, nuclear magnetic resonance, and XPS).However, this work underscores both the important role of the bottom interface in the charge transport through thiol-SAMs and the ability to use MEJs to elucidate details of metal-thiolate chemistry on surfaces.

Stability
The chemical stability of MEJs comprising SAMs, which depends on the interactions between binding groups and the bottom electrodes, is crucial for both fundamental research and practical applications.0][151][152][153] Annealing thiol-SAMs at a high temperature (e.g., >370 K) first results in the conversion of vacancies in the SAMs into large, ordered molecular domains (the aforementioned striped phase) before the eventual complete desorption of molecules from the surface. 154,155The occurrence of stages and their transition is rationalized by the competition between the reorganization and disruption of thiol-SAMs; that is, the molecular reorganization is dominant at a lower annealing temperature, and the desorption due to the chemical reaction of thiolates takes over as the temperature continues to rise.Notably, the annealing of thiol-SAMs also leads to the generation of disulfides, which is absent in the pristine SAMs and at lower temperatures (e.g., <370 K), implying a higher energy barrier for disulfide formation. 150The desorption temperature of disulfides (typically >450 K) is higher than that of thiolates, both of which depend the intermolecular van der Waals interactions as well as the surface coverage of SAMs. 155,156][160] The chemical stability of thiol-SAMs against oxidation is more difficult to affect because it is a property of sulfur-metal bonds and not the collective properties of the SAM.6][167][168] In SAMs of alkanethiols, the rate of oxidation evolves inversely with alkyl chain length and strongly correlates with the degree of order (indicated by water contact angles).Thiol-SAMs comprising alkyl chains <8 carbons are more susceptible to (photo-)oxidation, while chains >10 carbons facilitate the formation of more densely packed SAMs and therefore block active oxygen species from penetrating the monolayer to react with the thiolates at the metal interface.(It is reasonable to infer that this correlation between the oxidation rate of SAMs and the chain length can be generalized, e.g., in thiol-SAMs comprising oligophenylenes and oligoethylene glycols.)A few studies showed that the oxidation of thiol-SAMs also depends on the quality of substrates. 164,169,170The generation of disordered SAMs, which is a direct product of oxidation (be it in the dark or under UV irradiation), is dramatically faster on rough substrates (commonly found on directly evaporated metals) or polycrystalline substrates where the increased number of grains and grain boundaries undermine the stability of SAMs due to the varied binding strength of thiols and chemical reactivity. 171,172 Covalent/non-covalent thiol-free anchors In Sec.II A, we presented an overview of various aspects of thiols and their derivatives as some of the most extensively studied anchoring groups for molecular ensembles toward MEJs, including their advantages (e.g., the compatibility with a large variety of substrates) and limitations (e.g., the susceptibility against oxidation).As the demand for new materials and functions in both fundamental research and practical applications continues to grow, other anchoring groups are utilized in place of thiols and their derivatives to form molecular ensembles on various surfaces.

Silanes for SiO 2 /Si substrates
Silane anchoring groups were developed in 1983 to create oleophobic monolayers.Originally observed on oriented molecules physisorbed/chemisorbed onto hydrophilic surfaces, e.g., Pt, 173 silane monolayers resulted in materials with the thickness of a single molecule that are structurally equivalent or superior to Langmuir-Blodgett monolayers. 1746][177] Silanes (RSiX 3 , X ¼ Cl, OMe, OEt) are commonly used as anchoring groups that react with the silanol groups on the surface of clean SiO 2 /Si by silanization; [178][179][180][181] alternatively, organometallics (e.g., organolithium RLi and organomagnesium RMgX compounds) and alcohols (ROH) are used on SiO 2 /Si surfaces functionalized by halogenation and amination, which convert the surface Si-OH into Si-Cl or Si-NEt 2 . 182Studies on silane monolayers comprising -SiCl 3 -anchoring groups showed that a transition temperature governs the formation of monolayers from solution; e.g., silanization below the transition temperature yields monolayers in higher quality than those prepared from above. 183The transition temperature is an intrinsic property of the silane molecules, which is linearly dependent on the molecular length; i.e., the synthesis of silane monolayers comprising longer tails is more tolerant of higher temperatures than the shorter ones.The formation of silane monolayers starts with the physisorption of silane molecules onto the hydrated surface of SiO 2 /Si and the hydrolysis of -SiX 3 into -Si(OH) 3 .The trihydroxysilanes immediately react with the hydroxyl groups of the surface silanols to facilitate the covalent anchoring of the molecules.Similar to thiol-SAMs, the initial, disordered monolayers eventually convert into compact structures as a result of the increased surface coverage of molecules.Although that initial physisorption step is a self-assembly process, the subsequent formation of covalent bonds arrests selfassembly, leading to more robust monolayers than thiol-SAMs.This distinction is nontrivial in the context of MEJs because the ability of SAMs to continue undergoing self-assembly can be critical to their function, as is discussed in Sec.III E. In addition to solution processing, silane monolayers comprising short tails and high vapor pressures can be fabricated by evaporating the molecules guided by a flow of inert gas at room temperature or thermal evaporation in vacuum. 176,177,184,185Unlike thiol-SAMs, the covalent bonds formed by silanes are irreversible and do not result in a mobile adlayer.While the initial physisorption can be characterized as self-assembly, once covalent bonds form, the monolayers can no longer undergo self-assembly and cannot undergo in-place exchange or self-healing.However, silane monolayers do not undergo the chemical degradation previously discussed in Sec.II A 3 for thiol-SAMs.
The formation of robust silane monolayers (as well as monolayers using other aforementioned anchoring groups) on SiO 2 /Si substrates enables the investigation of their charge-transport properties in MEJs using a temporary, nondamaging top electrode (e.g., liquid Hg contacts) 186,187 and the direct deposition of metallic contacts via thermal evaporation 176,188 or soft lithography, 189,190 which is especially attractive because it allows the characterization of charge transport through the monolayers in packaged, integrated circuits.The rational synthesis of molecules and the selection of electrode materials (e.g., doping of the silicon substrate) work together to fine-tune the alignment of energy levels in MEJs comprising silane monolayers, resulting in molecular devices of particular interest, e.g., molecular diodes 175,191 and monolayer field-effect transistors (SAM-FETs) [192][193][194] that exploit the out-of-plane and in-plane charge-transport properties of monolayers, respectively (Fig. 6).In the case of molecular diodes, the silane molecules comprise a saturated alkyl spacer (r-bridge) and a p-conjugated moiety, and the monolayers grown on n-type Si substrates are contacted by 1 Â 10 4 lm 2 Al top electrodes with a similar work function fabricated by vacuum deposition.The rectification in these r-p monolayers is ascribed to the resonant tunneling through the frontier orbitals in the p-conjugated moiety of the molecules, while the bottom interface between the SiO 2 /Si and the anchoring groups, together with the r-bridge, contribute to the potential drops across the junction.In the case of SAM-FETs, the polarized bottom interface contributes to the dielectric behaviors of the silane monolayers as well as the elimination of trap states, allowing a reduced operating voltage and power dissipation on the devices and sustaining relatively high gate fields.][197] Monolayers formed on SiO 2 /Si are resistive to degradation over a broad range of temperatures, the treatment of acids, as well as shelf storage over a long period of time. 178,198,199Silane monolayers bearing alkyl chains are stable up to 350 C, which was found to be independent of molecular lengths, and incubation at higher temperatures leads to desorption/evaporation of molecules from the substrate due to the cleavage of C-C bonds (the cleavage of Si-O bonds happens at >725 C).In contrast to their stability against acids (e.g., 0.1 N HCl), the deterioration of the monolayers is pronounced upon exposure to aqueous bases (e.g., 0.1 N NaOH), which results in the hydrolysis of Si-O bonds and the complete removal of the molecules.The fabrication of monolayers, on the other hand, requires more stringent controls over conditions due to the reactivity of the anchoring groups (e.g., RSiCl 3 groups are easily hydrolyzed in ambient air) and the passivation of substrate surfaces.Recent work has revealed the thermodynamic phase-like transition in silane monolayers on SiO 2 and the critical temperature that dominates their quality. 200A temperature higher than the transition point threshold guarantees the formation of high-quality monolayers, while the effect of a lower temperature can be compensated by controlling reaction conditions, e.g., molecule concentration.

Carbon substrates
Molecular monolayers and multilayers electrochemically grafted onto carbon electrodes to form carbon monolayers are an alternative approach to serve the purpose of forming stable, practical MEJs.2][203][204][205][206][207] Carbon monolayers are not self-assembled for the same reasons as silane monolayers, which impart many of the same advantages and trade-offs as well as some that are specific to carbon monolayers.Their fabrication typically starts with the electroreduction of diazonium salts of the target molecules in solution at the surface of the carbon substrate, which generates radicals that rapidly react with the substrate to form covalent bonds.An additional layer comprising different functional molecules can be sequentially installed on top of the preceding layer via electrochemical grafting. 208,209The surface coverage of molecules shows a dependence on the material of the bottom electrode; e.g., the surface coverage of electrochemically grafted 4nitrophenyl diazonium is four times higher on glassy carbon than the basal plane of HOPG, which is rationalized by their different roughness and the number of available reaction sites (planes vs edges). 201he formation of C-C bonds between the molecules and the substrates is irreversible and stable against temperatures >500 C; therefore, phenomena such as surface reconstruction and molecule detachment that can undermine the structural integrity of the molecular ensembles are eliminated under conventional operating conditions. 210The symmetry and low dipole moment across the bonds between the p systems in both molecules and substrates are likely to reduce the barrier of charge injection at the bottom interface, significantly improving the electrical conductivity of the junctions.The growth process of carbon monolayers also naturally eliminates pinholes because the kinetics of the electroreduction of diazonium salts is fastest at the interface with the bottom electrode; however, the process is not self-limiting, and further reactions can occur at the carbon monolayer, particularly at the edge sites of the substrates, leading to the formation of multilayers thinner than 20 nm. 211,212These electrochemically grafted monolayers and multilayers are, in general, more compatible with vapor-deposited electrodes and conventional lithographic techniques than other types of monolayers (particularly thiol-SAMs) and exhibit highly reproducible electrical properties in laboratory protodevices and commercially available products 209,[213][214][215][216][217] (an example is shown in Fig. 7).
0][221] Due to the strong preresonance Raman scattering of 514.5-nm light, nitroazobenzene is a useful probe for studying binding between the molecules and various substrates.The nitroazobenzene radicals, generated from the electroreduction of nitroazobenzene diazonium salts, react primarily to the edge plane of PPF and the atomically flat basal plane of HOPG, while the latter is prone to generating undesired multilayers due to a higher population of defects.The determination of film roughness and thickness, by measuring the depth of trenches created by scratching the monolayers with an AFM tip, is particularly useful for verifying the successful deposition of target molecules.For example, the roughness increases from 0.155 nm for bare PPF surface to 0.236 nm for biphenyl monolayers and 0.504 nm for nitrophenyl and for nitroazobenzene monolayers. 211Raman and FTIR spectroscopy reveal varied, but non-zero tilt angles that further verify the chemisorption of molecules; physisorbed molecules typically lie flat on the substrate.

Emerging anchoring chemistry
Carboxylic acids and alkynes are often used as anchoring groups in place of silanes and thiols, respectively, to form SAMs on the surface of oxides and coinage metals, yet the influence of their bottom interfaces in the charge-transport properties of MEJs are rarely addressed in the literature.A study by Fracasso et al. 222 of the self-assembly of alkynes on Au and Ag revealed very little difference between MEJs comprising alkyne-SAMs and thiol-SAMs in terms of b and yields of working junctions.A subsequent study by Bowers et al. 223 compared the rates of tunneling transport through SAMs formed by alkynes on Au, carboxylates on Ag bearing a native AgO layer, and thiolates on Au or Ag to try to elucidate the influences of these anchoring groups in MEJs.Despite the drastically varied strengths of binding between these anchoring groups and the substrates, a comparison of the tunneling currents measured in MEJs did not show a distinguishable difference in their charge-transport properties, particularly tunneling decay coefficient b and injection current J 0 (Fig. 8).5][226] A possible explanation is that the large tunneling barriers created by alkanes mask subtle effects at the electrode interface(s).Measurements on SAMs bearing p-conjugated backbones (which create smaller barriers) may reveal these differences in the future.A recent study by Gu et al. 73 on carboxylate-SAMs approached the optimization of electronic coupling to the bottom interface from a different angle.In this work, carboxylate-SAMs were fabricated on a bimetallic electrode with a monatomic Ag (or Cu) adlayer on Au formed from underpotential deposition. 227,228The electrical conductance of carboxylate-SAMs on the adlayers is 30-40 times higher than those on bare Au, which is ascribed to a better degree of energy level alignment.These observations are rationalized by the splitting of surface molecular states on the adlayer-modified Au electrode into bonding and antibonding levels; the latter is closer to E f of the bulk electrode, reducing the tunneling barrier compared to bare metals.
While alkyne-and carboxylate-SAMs mimic many of the properties of thiol-SAMs, including their fragility, Crudden et al. 229 and Wang et al. 230 found that N-heterocyclic carbenes (NHCs) form ultrastable SAMs on Au.2][233] NHC-SAMs can be prepared by immersing either the bare Au substrate or an Au substrate passivated by thiol-SAMs into a solution of target molecules.The latter method facilitates the irreversible replacement of sulfides with NHCs, implying the replacement of thiolate-Au bonds with stronger carbene-Au bonds to form SAMs with high enough surface coverage to prevent the penetration of adventitious thiols in operando [Fig.9(a) and 9(b)].Notably, the ability to replace thiol-SAMs with NHCs depends on the size of the molecules; bulkier NHCs are only partially or completely ineffective at replacing thiols compared to complete removal by their smaller counterparts.SAMs comprising smaller NHCs also exhibit better stability at high temperatures (66 C-100 C) against oxidation or acidic/basic conditions, which is in sharp contrast to SAMs comprising bulkier NHCs.It is rationalized that the stability of NHC-SAMs is closely related to the ordering of the moleculessmaller NHCs more readily form densely packed monolayers that are less prone to structural deterioration than the bulkier ones.Furthermore, because of their stability, the derivatizations of the NHC-SAMs are achieved by simple aromatic substitution and S N 2 chemistry in situ, which enables the installation of functional groups such as ferrocene and dextran for biosensing. 234Recently Kang and colleagues 29,235  Thermoelectric characterization of the junctions shows the dependence of thermopower on the molecular length and a positive Seebeck coefficient [Fig.9(e)], which confirms that tunneling transport is hole assisted, i.e., the HOMO is closer to E f than the LUMO.
Several anchoring groups have been reported to facilitate the formation of SAM on graphene via non-covalent interactions with the aim of preserving the chemical structure of graphene. 236The noncovalent interaction between amines and graphene leads to the formation of SAMs that do not perturb the sp 2 hybridization in the graphene carbon network. 44,237Similar to the metal-thiolate adlayer in thiol-SAMs, the reversible nature of amine-graphene interactions also ensures sufficient molecular mobility on the substrate to generate densely packed SAMs of high surface coverage and junction stability. 238,239MEJs comprising SAMs of alkylamines show comparable electrical characteristics to thiol-SAMs, including values of b, but have significantly longer shelf stability (>30 days) than thiol-SAMs (typically 1-2 days) and a factor of 10 4 lower resistance in MEJs.Ferrocenes can also be used as anchoring groups to form SAMs on graphene via non-covalent p-p interactions 240 (Fig. 10).Similar to the well-established observation of rectification when ferrocenes are present at the interface with mechanically applied top contacts, ferrocene-graphene interactions induce rectification but are strong enough to drive the formation of SAMs while also mitigating the broadening of frontier orbitals, which localize them on the ferrocene moiety.As a result, the polarity of rectification in MEJs comprising ferrocene-graphene-SAMs is reversed with respect to ferrocene-terminated thiol-SAMs on coinage metals. 24Though it is not clear whether the mobility of ferrocene-SAMs is lower compared to amine-SAMs on graphene, the results in this work nonetheless suggest that the molecules are capable of forming densely packed monolayers that are stable over continuous bias sweeps.
Fullerenes, a family of carbon allotropes in which atoms are connected by alternating single and double bonds into fused rings to form a (partially) closed mesh, have been studied extensively as the active layer in organic electronic devices, e.g., in organic photovoltaics (OPVs) and organic field-effect transistors (OFETs).2][243][244] In a recent study, Qiu et al. 170 utilized these fulleroid-metal interactions to anchor PTEG-1, a fullerene derivative bearing glycol ether (GE) pendant groups, to metal substrates to form self-assembled bilayers (SABs) as depicted in Fig. 11.The topmost layer of PTEG-1 was found to undergo in-place exchange with arbitrary GE moieties to form SAMs that are anchored through non-covalent interactions with PTEG-1 chemisorbed onto metallic substrates (e.g., Au and Pt).These fullerene-SAMs retain the self-assembly properties of thiol-SAMs but form MEJs that are more mechanically robust and stable in air without encapsulation.Charge transport across these MEJs is nearly identical to their thiolate counterparts, producing identical values of b and J 0 as their alkanethiol counterparts and forming mixed monolayers of spiropyrans that exhibit light-driven conductance switching.The results in this work suggest that the entirety of the C 60 fullerene cage and the interdigitated GE chains serves as an equivalent charge tunneling/injection barrier to the S-Au interface in junctions comprising thiol-SAMs and that the GE phase is an efficient tunneling medium, as has been shown in thiol-SAMs. 245,246A subsequent study by the same authors showed that asymmetric charge transport across the MEJs comprising fullerene-SAMs with different top layers is mediated by the frontier molecular orbitals (i.e., HOMO and LUMO) of the molecules in the top layer and is unaffected by the fullerene anchoring group or the embedded dipoles arising from the polar C-O bonds in the GE chains at the bottom interface. 247The non-covalent nature of the interdigitated GE chains also enables in operando modulation of rectification in reconfigurable MEJs comprising fullerene-SAMs that rectify current.Their incorporation into microfluidic devices that encode stochastic data in chemical packets was proposed as a form of solid-state memory suitable for stochastic computation, which processes information as probabilities instead of exact sequences of bits.

III. THE MOLECULAR LAYER
The putative technological aim of ME is to incorporate molecules into electronic devices as core components that provide functionality arising from their chemoelectronic properties.The molecular layer, in combination with the top (Sec.IV) and bottom electrodes (Sec.II), dictates the behavior of tunneling electrons. 20Synthetic modification can impart interfacial dipoles, active chemical response, and a variety of electrical functions simultaneously working toward technological application and refining the deterministic manipulation of the behavior of charges traversing MEJs.In addition to the anchoring groups discussed in Sec.II, the bottom electrode-molecule interface can affect the electronic structure in the molecular layer of MEJs, for example, through aliphatic bridges that decouple molecular (p) states from the electrode(s) or the induction of interfacial dipoles that shift the vacuum level. 99In this section, we focus on these and other influences on charge transport in MEJs that arise from the properties of the molecules within.
Ensemble of molecules, via individual or collective effects, 48,49,248 acts as a scattering matrix for electrons impinging on the tunneling barrier that is defined by the molecular electron density states.Tunneling transmission probability is governed by the interference of electrons with these molecular states.This so-called quantum interference can be constructive or destructive, creating features in the transmission spectra. 530][251][252] Different modes of tunneling, such as inelastic, coherent, or non-coherent sequential tunneling, have also been observed in different MEJs. 253Inelastic tunneling has been utilized to study plasmon excitation in MEJs comprising SAMs of rigid molecules. 254Another useful tool for MEJs is the use of mixed monolayers to amplify the effects of molecular charge transport and properties, such that switching and rectification have seen significant advances in recent years. 25,29,142,255,256This section discusses several (organic) molecules that have been used to form self-assembled monolayers in (large-area) MEJs, categorized into four different classes as shown in Fig. 12, and ends with a discussion of the mixed-monolayer approach that combines different molecules in the same monolayers (Sec.III E).

A. Aliphatic molecules
Earlier research on MEJs was almost exclusively on SAMs of saturated alkanethiol molecules because of their ability to self-assemble in very ordered structures on surfaces.These studies established a benchmark against which to compare phenomena such as calculation of tunneling attenuation with molecular lengths (b), 35,36,170,257 functionalization or passivation of surfaces, investigating effects of molecular dipoles, 146,248,258,259 and involvement of defects 41 in tunneling charge transport (Fig. 13).
The benchmark methodology to characterize alkanethiol SAMs is the determination of b from the conductance of MEJs comprising various lengths of alkanethiols.On several coinage metal substrates, such as Au, Ag, and Pt, and in combination with several top electrodes, such as EGaIn, conducting probe (CP)-AFM, and polydimethylsiloxane (PDMS) micropores, a consensus range of values of b from the Simmons equation (Eq. 1) between 0.8 A ˚À1 and 1.2 A ˚À1 has been reached across several studies 35,36,42,44,95,170,257 [Figs.13(a) and 13(c)].This consensus extends to alkanethiols terminated with different anchoring groups 170 and/or modified with different functional groups.Further work has also investigated tunneling characteristics, including b from SAMs of oligoglycines, 260 oligoprolines, 261 and oligoglycols. 160,246Baghbanzadeh et al. 246 showed that MEJs comprising SAMs of oligo(ethylene glycol) produce small values of b, which was ascribed to hole tunneling by superexchange between the lone pairs of the oxygen atoms.
Odd-even effects in alkanethiols, commonly observed in their physical properties, 262 are also reflected in tunneling characteristics of these molecules, where odd-and even-numbered carbon chains can follow two separate trends in MEJ 35,42,44,45,263 [Fig.13(c)].More   Copyright 2019 American Chemical Society. 265WF, work function.
recently, Belding et al. 264 conducted an in-depth systematic study of alkanethiol molecules containing secondary amines and symmetrical tertiary amines to control for the molecular conformation in the SAM while keeping the electronic properties unaffected.Their work highlights the effect of molecular ordering on charge-transport properties of SAMs and shows that disorder can result in an artificial increase in current density.By systematically varying the alkyl substitutions via a process of elimination, the authors demonstrated that charge transport can be affected without varying the SAM thickness but by altering the molecular conformation because transport follows the molecular backbone.Finally, a higher dipole-induced rectification ratio, which was also demonstrated earlier by the same research group 265 [Fig.13(e)], were shown to be affected by better ordering of the SAM. Bruce et al. 259 showed that fluorination at the terminal and in the core of SAMs can disparately affect tunneling transport because of the presence of embedded dipoles.It has also been shown that due to anchoring-group chemistry with the metal electrodes, induced dipoles at the interface can affect the work function of the metal, in turn affecting the energy level alignment. 257,259Ai et al. 266 showed that rectification can be switched on and off by varying the interfacial properties due to protonation/hydration of SAMs of carboxylic acid groupterminated alkanethiol SAMs at the SAM//EGaIn interface.There are disagreements in the literature as to whether functionalization at SAM//EGaIn interface dictates charge transport, with some studies claiming that it does 267 and others claiming that it does not. 225,268Sangeeth et al. 269 used impedance spectroscopy on SAMs of alkanethiols to deconvolute the contributions of SAM to tunneling charge transport from the contact resistance at the SAM//EGaIn interface.The authors investigated the importance of errors in measured contact areas, contact resistance, the role of defects, 41 and topography of the bottom electrode 270 and their effects on resistance of SAMs measured at zero bias.In their follow-up work, Wang et al. 267 studied SAMs of halogen-substituted undecanethiols on Ag TS (templatestripped Ag) substrates and EGaIn top electrodes using impedance spectroscopy [Fig.13(d)].Moving down the halogen series synthetically, the authors reported an increase in dielectric constant ( r ) from 2 to 7.9, accompanied by an increase in conductance of 10 3 and a reduction in contact resistance from 7 to 2.1.These results emphasize the effect that polarizability of head groups can have on charge transport through SAMs in MEJs.
Jin et al. 255 studied mixed monolayers of alkanethiols by varying the molecular length, SAM composition, and manner in which the two components formed the mixed monolayers, homogeneously or heterogeneously, and their effect on tunneling conductance.Kumar et al. 124 used XPS and contact-angle measurements to understand the influence of disulfide impurities in SAMs with thiol anchoring groups using derivatives of dithiothreitol (DTT).The authors investigated mixed monolayers of similar molecules and reported that physisorbed disulfides did not show any length dependence of tunneling current until after the reduction of disulfides via free thiols from solution, yielding a b value of 0.57 A ˚À1 for the latter.Sauter et al. 258 used XPS to investigate SAMs of alkanethiols containing ester groups that induce embedded dipoles in the middle of the SAMs, showing that the dipoles have a direct influence on the work function of the bottom electrode based on the orientation of the molecular dipole. 258Xie and colleagues 109,245 applied a single-level model to SAMs of alkanethiols and dithiols on different substrates (Au TS ; Ag TS ; Pt TS ) using a CP-AFM (see Sec. IV A for details) top electrode, finding that charge transport occurred via a nonresonant tunneling process.Using the tunneling current-voltage behavior of the MEJs, the authors could extract molecule-electrode coupling parameter (C) and energy offset ( h ) between the molecular levels and E f using this single-level model.It was shown that dithiols coupled more strongly to the top electrode than the monothiols and that the energy offset was independent of the changing molecular lengths.The authors also demonstrated that both C and h reduced from Ag TS to Au TS to Pt TS as a linear function of their respective work functions.Additionally, SAMs of substituted aliphatic molecules have been used to functionalize surfaces for triboelectric nanogenerators (TENG). 271Recently, Cheng et al. 272 used cationic thiol-based SAMs to effectively manipulate the work function of Au substrates, enabling efficient TENG output characteristic.In their previous work, the authors also used fluorinated aliphatic molecules with trichlorosilane termini group, which upon treatment, forms a robust SAM on a PDMS or Al surface, showing long-term stability and highest output performance of these SAM-based TENG devices. 273Wang et al. 274 used thiol and silane SAMs with electron-donating and -withdrawing groups to affect the polarity and the triboelectric charge due to created surface dipoles.

B. r-p systems
While aliphatic molecules provide a good benchmark platform for performing charge-transport studies and spectroscopic investigations, their wide HOMO-LUMO gap precludes dynamic responses to stimuli.Alkanethiol SAMs, except for their interfacial properties, are known to behave more similarly to vacuum tunneling gaps than conjugated systems.Fully p-conjugated systems (discussed in Sec.III C) are usually rigid molecular wires that form less densely packed SAMs with more disorder because of reduced degrees of freedom available.Molecules that include both an unsaturated fragment (r) and a conjugated fragment (p), however, can combine the useful features of both.The r-framework provides a flexible structure that improves flexibility for self-assembly while at the same time decoupling the p-states of the molecule from the density of states of the bottom electrode.The former is significant because better ordering means more control over the electronic and charge-transport behavior of SAMs.The latter is even more significant because Fermipinning of the p-states can reduce the contribution and sensitivity of pstates to the electric field in the MEJ and, hence, its influence on the tunneling electron. 72,74,95,191,275Decoupling of p-states results in SAMs behaving as pseudo-gas-phase systems in which the molecular states are more accessible than in fully conjugated systems, as the latter are pinned to the fermi level of the electrode.The asymmetry in the structure is also replicated in the asymmetry of the electronic properties of SAMs (dipoles, electron density, etc.) because the transmission probability becomes different for V > 0 and V < 0. Further, these r-p systems provide the advantage of exploring mixed monolayers of these molecules and saturated molecules, which is discussed later in this section as well as in Sec.III E. Thus, the focus of this section is on how these systems control charge transport by placing localized, accessible p-states inside asymmetric tunneling barriers.

Molecular rectifiers
One of the early demonstrations of interesting functionalities in a r-p system was rectification in ferrocene-terminated alkylthiol molecules on Ag TS substrates with EGaIn as a top electrode. 24ijhuis et al. 24 demonstrated that alkanethiols with a polarizable ferrocene moiety as a head group produced higher tunneling current at negative bias, with rectification ratios R ¼ JðÀVÞ JðþVÞ of % 100 at 61 V.The authors further demonstrated that rectification occurs because the low-lying HOMO level of the ferrocene head group comes in resonance at negative bias such that charges experience a tunneling barrier defined only by the alkane fragment.At positive bias, the HOMO is pulled out of resonance by the electric field, and the tunneling barrier becomes defined by the entire molecule; i.e., d is larger at positive bias than negative bias, resulting in larger currents at negative bias. 38,250,252The authors further demonstrated the role of the top van der Waals interface by examining alkyl tails of odd and even numbers of carbon atoms.Yuan et al. 250 further investigated the effect of the position of the ferrocene moiety along the length of the molecule on the magnitude and polarity of rectification, verifying the role of the states localized on the ferrocene in the rectification mechanism [Fig.14(b)].The involvement of redoxactive species like ferrocene should not be confused with carrier generation and diffusion in bulk (organic) semiconductors.The short-length scales allow these processes to occur very quickly.For example, Trasobares et al. 276 showed that MEJ diodes comprising ferrocene-based SAMs can operate up to 17 GHz.Garrigues et al. 277 applied a single-level model to the ferrocene derivatives studied by Yuan et al. to explain the temperature dependence, confirming experimental rectification ratios and further adding to the understanding of electrostatic potential profiles in MEJs and their effect on rectification ratios.Copyright 2019 American Chemical Society. 281oon et al. 26 studied several polycyclic aromatic hydrocarbons attached to alkanethiols that affect rectification in pure and mixed monolayers.Rectification in bipyridyl-terminated alkanethiolates (BiPy) was demonstrated at the positive bias, i.e., opposite to the observation made in ferrocene-alkylthiolates, in MEJs with EGaIn as the top electrode.The charge transport at negative bias is completely dominated by a tunneling mechanism, while at positive bias, it was hypothesized to be a combination of hopping and tunneling.Later, Kang et al. 275 found temperature dependence of the rectification in BiPy SAMs that revealed the involvement of a Marcus inverted regime in transport.In mixed monolayers of BiPy, Kong et al. 256 also investigated the effect of the molecule-electrode interface by changing the lengths of BiPy and the alkanethiol diluent.It was shown that the wettability and rectification behavior of BiPy mixed monolayers can be altered by varying the lengths, resulting in conductance governed by the topography of the SAM-EGaIn interface.More recently, Park et al. 278 investigated the rectification properties of SAMs of BiPy-MCl 2 complexes (where M ¼ Co or Cu) attached to undecanethiol molecules.The BiPy-CoCl 2 complex was reported to rectify with current (R þ ¼ 82.0), while the BiPy-CuCl 2 did not.The rectification behavior was explained using DFT simulations, which stated that for the Co complex, only LUMO comes into resonance for the V > 0, while for the Cu complex, HOMO and LUMO both come into resonance for V < 0 and V > 0, respectively.Cho et al. 279 also demonstrated the role of Stark effect in systems containing other polycyclic aromatic hydrocarbons, such as naphtyl, phenanthrenyl, anthracenyl, pyrenyl, and benzo[a]pyrenyl, terminated with alkanethiol tails.The authors postulated that accessible molecular levels on the p-fragment can induce rectification in a temperature-independent tunneling regime via the Stark effect.Similar to the proposed rectification mechanism of BiPy SAMs, Qiu et al. 25 showed that fullerene-terminated alkanethiolates also show rectification at positive bias.It was shown, with support from DFT calculations, that the low-lying LUMO level participates in charge transport, decreasing d by the width of the fullerene cage at positive bias.
Rectification in MEJ of r-p molecules with silicon bottom contacts and EGaIn tips as top contacts has been demonstrated by Lamport and colleagues 280,281 [Fig.14(d)], while rectification values as large as 150 were shown by Broadnax et al. 282 for ferrocene-alkylsilane SAMs.Lenfant et al. 191 also studied SAMs with different p-conjugated head groups attached to alkyl tails anchored to silicon substrates, showing varied rectification behaviors across the studied molecules.Using a trivial analytical model, the authors demonstrated that because of Fermi-level pinning, the rectification behavior of the molecules was insensitive to the molecular orbital position in an otherwise isolated system.SAM-based MEJ studies incorporating p-conjugated donoracceptor moieties with alkanethiol tails also have been shown to generate asymmetric current-voltage curves. 251,283Finally, Yuan et al. 284 studied SAMs of polychlorotriphenyl moiety terminated with alkanethiols in their stable neutral and radical states, showing that stable radical species can be incorporated in MEJs with tunneling as the dominant form of charge transport.

Molecular switches
Responsive molecules have been studied in SAMs to monitor and investigate their chemical or electronic response to external stimuli, such as the change in conformation upon optical excitation.Smaali et al. 285 incorporated fused azobenzene-thiophene moieties attached to butanethiol tails into SAM-based MEJs with gold bottom contacts and CP-AFM top contacts.A high on/off ratio of 1.5 Â 10 3 was reported, and the conformation switching was monitored using contact-angle measurements and changes in thickness using spectroscopic ellipsometry.The high and low conductance states correspond to the cis and trans forms of the immobilized azobenzene switches, respectively.More recently, Kumar et al. 142 studied SAMs of spiropyrans with saturated alkyl ester chain anchored to the metal substrate.Upon illumination with UV light, the mixed monolayers of spiropyran in hexanethiol show a 35-fold increase in tunneling currents as the molecule conformation changes from spiropyran to merocyanine form.Unlike azobenzenes, the switching process does not affect d; rather, the push-pull nature of the merocyanine form shifts and delocalizes the unoccupied molecular states.The authors further showed that the fatigue in the reversible photoswitching could be reduced significantly for 100 cycles. 286The open merocyanine form can be also locked in the merocyanine form via protonation, which was then used to show the application of these devices as nonvolatile molecular memory with on/off ratios of % 10 3 . 287ecently, Han et al. 99 demonstrated responsive molecular switches based on methylviologen moiety with alkylthiol tails that show large on/off ratios % 10 3 and high rectification ratios % 10 4 [Fig.14(c)].The authors showed that the on/off and rectification ratios depended on the counter anion, and the effects are induced by the formation of viologen dimers and ion migration along the applied electric field.Schuster et al. 288 studied the photoisomerization of azobenzenesubstituted alkanethiol molecules (decorated with polar head groups such as H, CH 3 , and CF 3 ) between the cis and trans conformations and monitored the effect of changing molecular conformation on the electrostatics of the metal electrode.The authors showed that mixed monolayers of azobenzene derivatives with shorter alkanethiols effect larger changes in work function upon photoisomerization than pure SAMs of the same azobenzene derivatives, which also varied with polar or nonpolar head groups.These several examples of MEJs comprising ensembles of molecular switches where a change in electronic properties of the SAMs is driven by external stimuli underscore the usefulness of r-p systems in model ME devices; by decoupling the pstates from the electrodes, they can be manipulated by external stimuli, causing changes in conductance and rectification.

Miscellaneous
Zhang et al. 95 studied oligothiophene molecules with one to four thiophenes attached to butanethiol chains anchored to gold and silver substrates, with EGaIn and CP-AFM top electrodes showing nonlinear dependence of tunneling current on the molecular length [Fig.14(a)].Due to the decoupling of the p-conjugated thiophene units from the electrode by the r-bonded spacers, the height of the tunneling barrier decreased rapidly with increasing length, resulting in a reverse length dependence after bithiophene; i.e., J increased with d instead of decreasing.The quarterthiophene derivative was also used to demonstrate the mechanical and electrical robustness imparted to these SAMs by the flexibility of the butyl tails using an AFM top electrode. 94he nonlinear dependence of tunneling barrier was explained using the adapted two-barrier model, which was first used by Liao et al. 289 on SAMs of molecules containing alternating oligophenyl and alkyl fragments as two different tunneling barriers.The phenomenology of unusual length dependences has recently been reviewed elsewhere. 290

C. Fully conjugated molecules
For next-generation ME devices, the central idea aims to incorporate molecules that introduce varied functionality in an electronic circuit.The modulation and manipulation of current in molecular devices can be performed by varying either the molecular dimensions or their electronic properties, including frontier orbital gaps.As discussed in Secs.III A and III B, aliphatic molecules and r-p systems can be very useful to study structure-function relationships and electronic structure-based applications.However, the former act as an insulator while the latter (usually) create an asymmetric tunneling barrier.Another class of organic molecules that has been widely studied are the fully conjugated molecules.Unlike aliphatic and r-p systems, owing to their extended conjugation and (relatively) small frontier orbital gaps, these molecules create accessible energy states in MEJs.The tunneling probability is prominently affected in the experimental bias window by one or both frontier orbitals, which typically span the entire length of the p-conjugated molecules.Due to this, these molecular properties can be further used to affect and control charge transport by modifying these accessible states and shifting them in energy and/or varying (de)localization across the molecular backbone.We discuss recent advances in using conjugated molecular wires in MEJs split into the different functionalities that originate from the electronic properties of the molecules under study.

Quantum interference
Smaller frontier orbital gap (smaller tunneling barrier) from extended conjugation makes fully p-conjugated molecules more conducting compared to aliphatic systems.The high conductance of these conjugated molecules can be actively switched by utilizing the peculiarities of QI; in molecular junctions, destructive QI can cause the current to be suppressed by several orders of magnitude.The origin of destructive QI in molecular junctions is generally ascribed to the partial cancelation of the wave functions of tunneling electrons at certain energies.This type of interference is better understood classically as nodes in standing waves than as traversing waves canceling each other.A tunneling barrier defined by a molecule can be described as a scattering matrix composed of several molecular states that all contribute toward the total transmission (which is observed as electric current).The simplest system that would give rise to a perfect antiresonance QI dip [TðEÞ !0] is one in which the contribution of two molecular orbitals to conductance differs by a phase of p.This scenario naturally becomes more complicated when the combination of all the involved eigenstates are taken into account.Although several molecular orbitals contribute to a transmission peak, generally speaking, those in the vicinity of the peak in an energy landscape will contribute the most.For instance, resonance peaks in transmission spectra, where the TðEÞ ! 1, are usually located in the vicinity of the energies of molecular levels.Similarly, the destructive QI feature, which is usually identified in the form of a sharp antiresonance dip in the transmission spectra [TðEÞ !0], is usually located around the energy levels that contribute the most toward its origin.Destructive QI in ME has been thoroughly demonstrated in SMJs, providing fundamental insight, 291 but it is unclear how these results will translate to MEJs because the molecular states involved are very sensitive to conformation.
Although r-QI has only been observed in SMJs, 292,293 the rigidity of p-conjugated molecules facilitates the study of p-QI in MEJs.In fact, the first experimental evidence for destructive QI in ME was reported in MEJs.Fracasso et al. 53 observed the suppression of current in MEJs by synthetically manipulating the bond topology of pconjugated anthraquinone-based (AQ) molecular wires, showing that cross-conjugation suppressed current even more than interrupted conjugation.Carlotti et al. 51 discovered through-space QI in fully conjugated molecular wires in which transport occurs between two phenyl rings oriented either face-on or edge-on.In the latter case, the two phenyl rings were forced into coplanarity in SAMs, and the same molecules, measured in SMJs, showed the opposite behavior, becoming more conductive when stretched between electrodes. 294Jia et al. 68 incorporated SAMs of the through-space molecular wires into transistors, demonstrating the gating of destructive QI features.Famili et al. 295 used the same strategy to gate QI effects in MEJs arising from bond topology [Fig.15(b)].Carlotti et al. 296 further explored many derivatives of AQ molecular wires, demonstrating that the degree of current suppression can be controlled by including electron-donating and -withdrawing functional groups.They found that the most electron-poor AQ wire could be switched between linear and crossconjugated states with applied bias, demonstrating a QI-based, twoterminal molecular memory device 297 [Fig.15(a)].Zhang et al. 298 further demonstrated control over the positions of destructive QI features in the transmission spectrum through a combination of bond topology and electronegativity.Liu et al. 39 found evidence of intermolecular QI effects that are present in MEJs but that can also be observed in SMJs by trapping two molecules in the junction.Recently, Soni et al. 299 demonstrated that the involvement of intramolecular pathways can be used to manipulate the magnitude of tunneling currents in both SMJs and MEJs by placing cross-conjugated pathways in parallel with linearly conjugated pathways in fluorenone-based molecular wires.The experimental observations were supported by a superposition model that accounts for conductance through parallel pathways in a quantum mechanical circuit that was proposed for SMJs by Vazquez et al. 300

Transport over long molecules
In quantum mechanical tunneling, the tunneling probability across a tunneling junction decays exponentially with the width of the tunneling barrier (d), as expressed by the Simmons equation (Eq.1).This exponential decay precludes the fabrication of tunneling junctions with electrode separations of more than a few nanometers.][305][306][307] Transport can also become thermally activated, at which point length dependence of the charge-transport mechanism becomes insignificant. 306,307Smith et al. 308 studied SAMs of short and long thiophene oligomers grown on metal substrates by click chemistry.Their measurements revealed a temperature-independent charge-transport mechanism on short oligomers that evolved into a temperaturedependent hopping mechanism in longer oligomers.Later, Taherinia et al. 309 measured the conductance of a 4-nm-long molecular wire containing aromatic and aliphatic six-membered rings distributed at different positions along the molecular length using a Au AFM top electrode.Owing to the length of the molecular wire, conductance was measured in the hopping regime.The conductance was reduced with the synthetic addition of the aliphatic rings (reducing the extent of pconjugation) independent of their position in the wire.

Embedded dipoles
Considerable effort has been put into manipulating energy levels at interfaces using embedded dipoles in molecular monolayers that carry over into the manipulation of tunneling barriers in MEJs.Embedded dipoles alter the shape of the tunneling barrier at the position where the induced dipole is located.G€ artner et al. 310 grew SAMs of conjugated molecules containing molecules with differently oriented nonpolar phenyl and polar pyrimidine groups with respect to the thiol anchoring group.Using photoelectron spectroscopy, the authors demonstrated that depending on the position of the dipoles, the work function of metal can vary by % 0.9 eV.Kovalchuk and colleagues 49,50 examined SAMs of terphenyl derivatives containing pyrimidyl rings in their cores with dipoles oriented toward and away from the bottom substrate that altered charge transport by altering the work function, causing asymmetric conduction that depended on the relative polarities of the dipoles and the applied bias.This was further supported by theoretical work by Abu-Husein et al. 48and spectroscopic studies by Sauter et al. 248 W€ achter and colleagues 311,312 also studied electron transfer dynamics in functionalized heteroaromatic molecules using XPS.These collective effects, arising from individual molecular properties affecting macroscopic tunneling current, show promising avenues for the use of conjugated molecules in ME devices.However, some studies by Chen et al. 226 302 substituted by halogens at the SAM//EGaIn interface showed an insensitivity of tunneling current.The dielectric behavior and tunneling rates were hypothesized to be insensitive as the modulations in the electrostatic potential due to the spatially localized, electronegative halogens, suggesting an interplay among dipoles, conjugation, and polarizability that can be exploited for fine-tuning the electrostatic profile of MEJs.

that utilized SAMs of biphenyl
G€ artner et al. 310 used XPS, water contact angle, Kelvin probe, and DFT to study SAMs of heteroaromatic molecules containing nonpolar phenyl and polar pyrimidine groups incorporated in two opposite directions.They showed that while the physical characteristics of the SAMs, such as tilt, packing density, and wetting, were independent of dipole orientation, the latter shifted the work function of Au by $0.9 eV.Studies of terphenyl molecules on Ag substrates by Sauter et al. 248 also showed the same behavior of shifting the work function of the metal based on the dipole orientation of the SAM.Asyuda and colleagues 302,313 showed that changing the composition of binary SAMs of conjugated biphenylthiol molecules by increasing the percentage of electronegative F atoms at the SAM//EGaIn interface with respect to H atoms (or CF 3 and CH 3 groups at the interface) also systematically changes the work function due to embedded dipole and, subsequently, the tunneling charge transport for these SAMs [Fig.15(d)].These varied observations emphasize that molecular design has complex, nontrivial effects on the electrostatic profiles of MEJs and SAMs, further underscoring the central feature of ME: Electrical properties respond directly to subtle changes to molecular structure and conformation.

Thermoelectricity
Thermoelectric measurements on MEJs are becoming more common in ME, but there are still only a handful of experimental studies on SMJs and even fewer on MEJs.We briefly discuss the recent advances in the field and direct the reader to a recent review on the topic by Park et al. 314 for a more thorough treatment of the topic.Reddy et al. 315 reported the first measurement of Seebeck coefficients in SMJs.They related the Seebeck coefficient S junction to the transmission function T(E) using Eq. 4, where the magnitude of S junction depends on the slope of T(E) at E f .Experimentally, S junction is usually determined using the differential method by measuring the ratio of the voltage difference DV and temperature difference DT according to Eq. 5. Thus, resonances caused by molecular orbitals and QI features (because they are sharp) near E f should be observable as large values of DV, and pconjugated molecules are more likely to produce large values of S junction (because their frontier orbitals tend to be close to E f ).Tan and colleagues 316,317 investigated the thermoelectric properties of SAMs of conjugated oligophenylenes using a thermal-AFM top electrode.The authors reported positive values of the Seebeck coefficient that increased with the increasing molecular length for both mono-and dithiol-terminated molecules, while negative values were reported for molecules with isocyanide anchoring groups.This result suggests that transport is dominated by the HOMO in the former case and LUMO in the latter, as the sign of the Seebeck coefficient depends on the sign of the charge carrier.This observation was also supported by simulations that showed HOMO and LUMO being closer to E f for the thioland isocyanide-terminated molecules, respectively.
Park and Yoon 318 used EGaIn as the top electrode for measuring the thermoelectric properties of oligophenylenes.By measuring the Seebeck coefficient of the Ga 2 O 3 layer, the authors confirmed that Ga 2 O 3 did not influence the measurements and that consistent Seebeck values were obtained in comparison to SMJs. 315This was the first experimental measurement of S junction in a large-area MEJ, i.e., with a macroscopic top contact rather than an AFM or STM tip.In their follow-up work, Park et al. 319 performed these measurements on SAMs of alkanethiols on Au substrates, which showed that thermopower decreased with increasing molecular length.Further, the authors measured power factors P ¼ GS 2 of SAMs of alkanethiolates and oligophenylenethiolates with varying length. 319They also introduced a parametric semiempirical equation to study the length dependence of the power factor.Recently, Ismael et al. 320 measured anthracene-based molecular wires, with phenylacetylene linkers and different anchoring groups, using a CP-AFM top contact.The authors observed negative values of S junction for molecules with S-CH 3 and pyridyl anchors and a 10-fold difference in magnitude, providing a way to fine-tune the thermopower via molecular design.Further, they found positive values of S junction for molecules with thioacetate anchoring groups, which was explained using DFT simulations predicting HOMO to be closer to the Fermi level for the thioacetate molecule while LUMO dominates charge transport for the rest of the molecules.Wang et al. 321 further measured SAMs of similar anthracene-based molecular wires but varied the connectivity of the linkers to the core using a modified CP-AFM setup.The thermoelectric power and conductance of these SAMs were measured while varying the tilt angle of SAMs by applying different load forces from the AFM tip.With increasing load force, S junction decreased, while the electrical conductivity G increased.The power factor was shown to go through a maximum with the load force; hence, the tilt angle was increased, providing another way to fine-tune the thermoelectric properties of MEJs.The MEJ-based thermoelectric studies discussed in this section primarily involve conjugated molecules because smaller frontier molecular orbital gaps tend to lead to larger thermopowers.However, Cho et al. 322 showed recently that the thermopower of saturated molecules can be enhanced using superexchange coupling in SAMs of oligo(ethylene glycol) molecules.

Miscellaneous
Lenfant et al. 323 demonstrated that the conductance of SAMs can be manipulated via the photochemical isomerization of azobenzene units in a lateral loop as part of a fully linear-conjugated molecule in an MEJ.The conductance switching ratio was shown to be as high as 10 3 .Xie and colleagues 301,324 applied a single-level model on SAMs of oligophenylenethiols and dithiols contacted by a CP-AFM top electrode [Fig.15(c)].Using experimentally obtained parameters, such as transition voltages (V trans ) for negative and positive bias, tunneling currents, and energy-offsets between the HOMO and Fermi level (using ultraviolet photoelectron spectroscopy), the authors elucidated that the Fermi-level pinning of HOMO is the cause for the weak dependence of the tunneling barrier height on the work function of the metal.The authors showed higher molecule-electrode coupling for dithiols than monothiols and decreasing energy offset with molecular length and electrode work function.In a separate study, Wang et al. 325 studied both SAMs of aliphatic and aromatic molecules in EGaIn MEJs, showing that using inelastic tunneling electrons, surface plasmons at the gold electrode can be excited.The data suggested that intermittent light emission (blinking) could be controlled by the molecules in the SAMs.For aliphatic molecules, the conformation changes in the junction resulted in blinking over time, while the aromatic molecules with rigid backbone gave stable light emission.The authors were also able to estimate the effective electric contact area at the top interface by monitoring the emissive plasmon sources.De Nijs et al. 326 also utilized plasmonic tunneling junctions comprising SAMs contacted by Au nanoparticles to create a plasmonic cavity that can be used to identify different redox states using surface-enhanced Raman spectroscopy.The authors observed shifts in vibrational frequencies corresponding to the different redox states accessed due to the presence of hot charge carriers generated in the plasmonic hotspots.Although this last example is not of a large-area MEJ, it, along with all the studies discussed in this section, highlights the utility of conjugated molecules in tunneling junctions.

D. Biomolecular ensembles
Organic, organometallic, and inorganic molecules do not represent every class of molecules available for ME.Next-generation bio(nano)electronics is an active area of research in ME, as it offers the possibility of incorporating a vast library of electron-transporting biomolecules, such as protein complexes, homopeptides, and DNA assemblies in bottom-up MEJs. 327In MEJs, it offers not only the precision and complex function provided by biosynthesis but also the ability to define robust nanostructures through self-assembly.While several theoretical and SMJ studies have been performed on biomolecules in tunneling junctions, [328][329][330][331][332][333] experimental studies on MEJs facilitate the elucidation of charge-transport mechanisms, electronic properties, device stability, etc. 327 Unlike SMJs, the physical properties of biomolecules, such as self-assembly, can be utilized in static, solidstate devices comprising MEJs to control the orientation and selective binding to electrode surfaces and, consequently, the mode of electron transport across these bioelectronic devices. 327The long-term stability of the performance of bioelectronic devices comprising MEJs requires the optimization of the self-assembly of biomolecules, molecule-electrode interactions, and molecule-molecule interactions.While electron transport through biomolecules, such as proteins, has been thoroughly studied in aqueous solution, embedding them in solid-state devices is a relatively new development concomitant to advances in top contacts.In this section, we discuss some commonly studied biomolecules in MEJs.

Photosystem I
Photosystem I (PSI), or light-driven plastocyanin-ferredoxin oxidoreductase, is one of the two membrane-bound protein complexes in green plants, algae, and cyanobacteria that is responsible for photosynthetic reactions that convert light energy into spatially separated electron-hole pairs. 336The PSI electron-transport chain consists of a chlorophyll electron donor (P700) and acceptor (A 0 ); another quinone intermediate electron acceptor (A 1 ); and finally, three iron-sulfur-based centers (F X , F B , and F A ).The chlorophyll molecules act as antennas to funnel excitons into the electron-transport chain where the cascade of donors and acceptors generate holes and electrons on P700 and F A/B , respectively, combining many potentially interesting functions into monodisperse, nanoscale complexes that can be readily harvested from spinach leaves.Castañeda Ocampo et al. 80 studied Au surfaces passivated with PSI complexes whose orientation was dictated by director SAMs that modified the surface chemistry of the Au [Fig.16(c)].They observed a clear relationship between the average orientation of the protein complexes and asymmetry (rectification) in the tunneling current.PSI was oriented in the "up" and "down" direction, i.e., with P700 and F A/B closer to the substrate, respectively, by using director SAMs favoring either ionic or hydrogen-bonding interactions.Later, Gordiichuk et al. 337 used phage display to oriented PSI with near-perfect selectivity on transparent metal oxide surfaces.Qiu et al. 86 exploited the ability to orient PSI on Au to construct selfregenerating, soft, stretchable biophotovoltaic devices with EGaIn counter electrodes.Recently, L opez-Mart ınez et al. 335 studied electron transport through individual PSI complexes using electrochemical STM, suggesting that long-distance current is facilitated by PSI electrochemically gated around the redox potential of P700 [Fig.16(d)].One of the remarkable and still poorly understood properties of protein complexes is their ability to conduct tunneling charges over long distances.The short axis of PSI is % 10 nm, yet it supports activationless transport, with conductances usually found in conjugated molecules that are <3 nm in length.

Azurin
Azurin (Az) is a periplasmic blue copper protein 338 that moderates electron transport in the bacteria Pseudomonas aeruginosa, Bordetella, or Alcaligenes by undergoing redox between Cu(I) and Cu(II).The preservation of the structure and function of Az 339 after it self-assembles on surfaces makes it a promising candidate for biodevices, and several studies have been performed on Az-based solid-state protein devices, which we discuss here.
Chi and colleagues 340,341 studied the assembly of Az on a surface linked to the gold substrate via alkanethiol molecules and probed electron transport using an STM tip.Alessandrini et al. 342 used Az to form single-protein-based wet transistors exemplifying the role of electron transport in bionanodevices.Li et al. 343 used CP-AFM as the top electrode to investigate the monolayers of Az protein that exhibit electron transport across Az monolayers.The authors demonstrated that the conductance of monolayers can be manipulated to be temperature dependent or independent by varying the AFM load force (i.e., inducing structural change), which should be considered as an important factor when conducting electron transport experiments in bionanodevices.Yu et al. 253 performed inelastic tunneling spectroscopy (IETS) on Az protein in solid-state tunneling devices.They demonstrated phonon-coupled electron transport across MEJs, suggesting that Az exists in its native state also in solid-state MEJs by comparing the properties of Az in the solution state.Ruiz et al. 344 measured conductance of Az in single-protein metal contact in an electrolyte environment using electrochemical STM.The authors studied Az by mutating the docking site of blue Cu protein, changing the electron transport mechanism from the Cu-mediated two-step process to direct, coherent tunneling.Fereiro and colleagues 334,345 investigated charge transport in solid-state devices comprising Az monolayers sandwiched between metal electrodes [Fig.16(b)].Similar to the r-p systems described in Sec.III B, the authors used a hydrocarbon spacer between Az and one of the electrodes to weaken the electronic coupling between the protein and electrode.The other Au substrate was contacted by an Au-S bond to the protein, changing the transport mechanism from off-resonant tunneling (without spacer) to nearresonant tunneling (with spacer).
Recently, Fereiro and colleagues 346,347 further demonstrated a solid-state protein conductance switch that shows a difference in conductance based on the orientation of the Az protein.These results were supported by the work of Romero-Muñiz and colleagues 348,349 using DFT and molecular dynamics simulations on Az systems.The authors studied two mutants of Az that exhibited enhanced electron transport compared to a saturated organic molecule as shown in Fig. 16(a). 81Kayser et al. 350 studied charge transport in solid-state MEJs of Az proteins for temperatures as low as 4 K, revealing that charge transport is mediated by quantum tunneling and not hopping, despite the comparatively large tunneling junction defined by the protein complex.Finally, Mukhopadhyay et al. 351 reported a cross-laboratory comparative study across several research groups using different bioelectronic devices comprising MEJs of proteins, showing the preservation of the charge-transport mechanism even with varying electronic coupling.Published under an exclusive license by AIP Publishing

Multi-heme cytochrome
Heme cofactors (iron protoporphyrin IX) are one of the most pervasive cofactors to hemoglobin in nature.The c-type cytochromes represent one of the largest families of heme-containing proteins. 352ulti-heme cytochrome-c can contain multiple heme proteins, and the vinyl groups of iron protoporphyrin IX are attached to the two cysteine side chains that have been studied extensively in the context of their electronic properties.The near-linear arrangement of hemes in this protein is believed to facilitate electron conduction.
Garg et al. 353 measured transport across monolayers of multiheme cytochrome-c using symmetric gold contacts and discovered that electron transport across the heme-proteins was dominated by a coherent tunneling mechanism.The current was found to be temperature independent, suggesting tunneling transport as the primary mechanism.The conductance of multi-heme-proteins was found to be three orders of magnitude higher than single heme-and heme-free proteins, signifying the importance of multi-hemes in charge transport.Later on, Futera et al. 354 also demonstrated that the electron transport mechanism in 3-nm-long multi-heme protein bioelectronic MEJs is off-resonant, coherent tunneling.Their results supported temperature independence of the protein conductance in solid state-very different from the heme-to-heme electron hopping that occurs in aqueous conditions.

Peptides
Apart from proteins, there are also a handful of studies on peptide-based solid-state devices.Sepunaru et al. 355 studied electron transport through homopeptides as a function of their structural properties, such as length, secondary structure, and charge on its residue.They showed the electron transport to be temperature independent except for helical peptides, which exhibited very-low activation.Naaman and Waldeck 356 and Torres-Cavanillas et al. 357 showed that room-temperature chirality induced spin selectivity, which was also shown for metallopeptides containing a Tb 3þ metal center in helical lanthanide-binding peptides.

E. Mixed monolayers
Unlike other sections that focus on different classes of molecules used in MEJs, this section centers on the usefulness of mixed monolayers in which two or more different molecules are incorporated into a (self-assembled) monolayer (Fig. 12).As shown in Figs.17(a) and 17(b), mixed monolayers can be used (i) to dilute molecules with a large head group to anchor size/volume ratios on the surface by using other shorter diluent molecules [Fig.17(a)] or (ii) to mix two structurally similar molecules in different molar ratios while varying functional groups, orientation, anchoring group, etc., to affect the junction properties [Fig.17(b)].Mixed monolayers offer three key advantages over homogeneous monolayers: (i) They facilitate physical-organic studies to examine the effects of the supramolecular structure of a monolayer and heterogeneous surface structure/composition on charge transport across MEJs, (ii) they enable the modulation of the electronic function of MEJ via molecular dilution, and (iii) they overcome the limitations of homogeneous monolayers comprising molecules with bulky head groups whose function in MEJs can be inhibited by side reactions, steric crowding, and disorder. 20,88e first strategy, shown in Fig. 17(a), enables the formation of SAMs that are otherwise inhibited by the steric crowding of bulky head groups that cause vacancies between the anchoring groups and imperfect surface passivation.This strategy leverages the ability of SAMs (i.e., systems in the processes of self-assembly) to undergo inplace exchange in which a homogeneous SAM is immersed in a solution of a diluent molecule bearing the same anchoring group, shifting the equilibrium toward a SAM of the diluent.Since exchange is fastest at defects, immersion for a limited time can displace mispacked, disordered, bulky molecules and replace them with ordered diluent molecules.For instance, Kumar et al. 142 showed that exposing disordered SAMs of spiropyrans with alkyl thiol anchors to shorter alkanethiols led to mixed monolayers that showed fatigue-resistant, larger switching ratios upon illumination compared to the pure SAMs of spiropyrans.Mixed monolayers can also accommodate molecules with competing binding groups by pre-passivating a surface with a SAM.Qiu et al. 25 embedded alkanethiols terminated with bulky fullerene cages in SAMs of alkanethiols to form a mixed monolayer (shown in Fig. 5), which rectifies current due to the proximity of the fullerene cage to the top electrode.Similarly, Kong and colleagues 256,358 investigated the rectification ratios of mixed monolayers of BiPy-terminated alkanethiols diluted with alkanethiols of varying length, which in turn affected the molecule-electrode interface by altering the surface wettability.The same authors recently demonstrated that while interstitial mixed monolayers could be used to enhance the stability of MEJs, 359 the dilution of rectifying molecules with nonpolar molecules can also affect the electron-hopping process in the Marcus inverted regime. 29chuster et al. 288 used mixed monolayers of azobenzene photoswitches substituted with different head groups that affected the work function of the substrate via reorientation of molecular dipole due to photoisomerization.In their study on PTEG-1 systems as an alternative for thiol-based SAMs (Fig. 11), Qiu et al. 170 used mixed SABs in which a monolayer of PTEG-1 molecules supported a second monolayer comprising a mix of short alkylglycolether chains separating the oligoethyleneglycol-functionalized spiropyrans on the surface.The authors used the mixed monolayer strategy to affect photoswitching on the surface, similar to the works of Kumar et al. on thiol SAMs, but with fullerene anchors instead of thiols.
The second strategy, shown in Fig. 17(b), enables the manipulation of the properties of MEJs that affect charge transport across the junctions.Jin et al. 255 demonstrated that by growing mixed SAMs of alkanethiols with different lengths, fine control can be achieved over the heterogeneity of the surface topography via mixed monolayers.They showed that the modality in the distribution of tunneling currents largely depended on the difference in the lengths of the alkanethiol molecules.Kumar et al. 124 used XPS and contact-angle measurements to investigate the charge-transport properties of mixed monolayers of alkanes with both thiol and disulfide anchors.Similar to the findings of Jiang et al. 136 in which leakage current in mixed monolayers of ferrocene-terminated alkanethiols with disulfides and thioacetate "impurities" resulted in different degrees of rectification, Kumar et al. 124 demonstrated that SAMs of pure disulfides with varying aliphatic chains did not exhibit length-dependent tunneling currents, while mixed monolayers grown from the same disulfides and then exposed to ethanethiol underwent metathesis at the gold interface, restoring the length dependence.Castañeda Ocampo et al. 80 showed that the magnitude and polarity of rectification responded to the degree and direction of the orientation of PSI complexes in mixed monolayers.Asyuda and colleagues grew mixed monolayers out of biphenyl molecules with thiol anchors but with head group substitutions that introduce oriented dipoles: CH 3 and CF 3 302 and H or F 313 [Fig.17(b)].By controlling the ratio of the two molecules on the surface, the authors showed a systematic shift in the work function, and hence, in the transition voltages and current density of these MEJs.

IV. THE TOP ELECTRODE
Before the application of a top contact, the bottom electrode and molecular layers are essentially free-standing molecular ensemblestypically thiol-SAMs-and can, therefore, be interrogated and understood as such.In this section, we discuss the fabrication of MEJs from these ensembles using state-of-the-art methods to solve the unique challenge of applying a top contact.We divide the methods of making a top contact into four parts: (1) solid metallic electrodes, i.e., applying solid-metal electrodes directly to a molecular ensemble supported by a conductive substrate; (2) liquid metallic electrodes; (3) hybrid electrodes, e.g., using a buffer layer between the molecular layer and the top contact; and (4) nonmetallic electrodes.There are many methods for attaching electrodes to molecules, and an entire review could be devoted to the topic. 17,360Here, the focus is on techniques that install a top contact to MEJs.We will not discuss techniques that form nanogaps, such as on-wire lithography, self-aligned lithography, and on-edge molecular junction, for which there are already excellent reviews. 21,361 Solid metal electrodes

Conducting probe atomic force microscopy
Originally developed by Wold and Frisbie 362 to form metal-molecule-metal junctions, CP-AFM, brings a metal-coated, conductive AFM tip into contact with a surface supporting a molecular ensemble.Like STM-BJs, CP-AFM forms transient junctions; Unlike STM-BJs, these junctions comprise hundreds of molecules and persist for as long as the tip is held in contact with the substrate. 363,364Alternatively, an AFM tip can be used to address preformed nanoscopic contacts such as a conductive nanoparticle predeposited on a SAM. 365Since the size of the nanoparticle is well defined, the number of molecules contacted can be determined from the packing density (which can be determined from several spectroscopic techniques); thus, the per-molecule conductance can be calculated.MEJs with CP-AFM top contacts are particularly well-suited for characterizing the nuances of bottom electrodes, interfaces, and molecular layers (i.e., Secs.II and III) because they can efficiently probe per-molecule conductance in small, defect-free areas of molecular ensembles, eliminating the influences of substrate topography and disorder that are endemic to large-area MEJs.Furthermore, the mechanical and the electrical properties of the molecule/substrate can be interrogated simultaneously. 94,343,366The force of the tip can also be controlled to optimize the moleculeelectrode interface and study the effects of deformations or rearrangements of the molecules. 362,367This fine control over the molecule interface is also ideally suited for studying the correlation between energy level alignments and the chemistry of the interface.For example, comparing aromatic thiols to aromatic isocyanides, 368 alkane monothiols to alkane dithiols, 109,369 and oligophenyl thiols to oligophenyl dithols. 301,324CP-AFM is a widely used technique, with a large body of work that has been reviewed elsewhere. 21Since the focus of this review is on MEJs with potential technological implications and MEJs formed using AFM tips are unlikely to find direct technological applications, they will not be discussed further.

Crossed wire junctions
Although crossed wires have been long used to form metallic tunneling junctions, 4 Gregory 370 improved on the methodology enough to form MEJs and perform IETS on them.Kushmerick et al. 371 further improved on the technique, measuring the I/V characteristics of MEJs comprising SAMs, as shown in Fig. 18.To fabricate a crossed wire junction, two metallic wires of % 10 lm in diameter are mounted in an orthogonal geometry.A SAM is grown on one of the wires, and it is positioned perpendicular to an applied magnetic field.A small current through this wire uses the Lorentz force to control its deflection from the other wire, allowing them to be brought into contact sufficiently gently to avoid damage to the SAM.As discussed in Secs.II A 3 and II B 3, SAMs are both chemically and mechanically fragile.Once an MEJ is formed between the wires, a voltage is applied to measure the conductance of the junction. 372,373Since different metallic wires can be used, the work functions of the electrodes and junction asymmetry can also be varied. 374Yoon et al. 375 reported a device-level crossed wire junction that enables the simultaneous measurement of I/V curves at variable temperatures while performing Raman and IETS measurements.Like CP-AFM, crossed wire junctions are transient but form small, well-defined junctions that are ideally suited to study the details of transport through SAMs of a wide variety of molecules, 371,[376][377][378] revealing the vibronic contributions to transport as well as the IETS selection rules. 373,374As a result of the fine control over the deflection of one of the wires, the contact force between the wires can be varied precisely, allowing the junction area to be varied, i.e., more molecules are contacted when the deflection current in the wire is increased. 376,377The number of molecules in a crossed wire MEJ is, however, difficult to estimate because the curvature is not well-defined and the orientation of the SAM with respect to the wires is difficult to determine exactly.Crossed wire junctions share many of the advantages and limitations (i.e., in the context of technological applications) of CP-AFM junctions and, thus, will not be discussed further for the same reasons.

Nanotransfer printing
Nanotransfer printing (nTP), first introduced by Loo et al., 379,380 is a technique to make MEJs with the SAMs acting as covalent "glue" to transfer thin metal films from a stamp, as shown in Fig. 19.Elastic stamps of PDMS fabricated by soft lithography are commonly used to transfer patterned electrodes by nTP. 381,382Rigid stamps can be made from GaAs by locally etching the GaAs with a patterned resist layer as an etching mask. 383The elastic properties of PDMS allow it to conform to small variations in height, mitigating the effects of substrate topography that can be detrimental to rigid stamps; however, that deformation can reduce the fidelity of the contact patterns.In both cases, the size and ratio of the pattern determines the number of molecules contacted.An important feature of nTP is the ability to fabricate multiple electrodes with different patterns and sizes simultaneously, which is useful both in the context of potential technological applications and for the statistical analyses that are described in Sec.I C. 384 Since the solid metallic films on the stamp are transferred in their entirety, the resulting structures do not suffer from short circuits induced by filaments, which is a common failure mode with vapordeposited metallic top electrodes. 385Suggestive of technological applications vis-a-vis integrated circuits, nTP has been used to fabricate crossbar arrays of MEJs in high yields. 386,387The crossbar architecture is key to the realization of memory and logic devices based on MEJs, [388][389][390][391] as it capitalizes on the ability to include multiple functions into a single junction 99 in high-throughput, low-cost parallel fabrication.The primary disadvantage of nTP is that it requires a mechanically robust molecular layer that covalently binds to the top contact with a thiol group, restricting the class of materials that can be used 386 and precluding the investigation of weak or physisorbed contacts.However, the only physical limitation is interfacial free energy of the stamp, which, as Niskala et al. 392 reported, can be reduced significantly by the choice of elastomer, resulting in high yields of stable junctions.The liftoff, float-on approach is inspired by the sample preparation for transmission electron microscopy experiments to deposit GaAs on other substrates. 393It was developed by Moons et al., 394 Vilan et al., 395 and Vilan and Cahen 396 as a soft deposition method to reduce electrical shorts in MEJs.The general procedure for the lowest in first out (LOFO) approach is illustrated in Fig. 20.In this method, a thin metallic film is deposited on a sacrificial substrate, lifted off using a detaching agent, and then floated on top of a conductive substrate supporting a molecular ensemble, thus resulting in an MEJ in a sandwich configuration.Similar to nTP, LOFO avoids filament formation from vapor deposition and avoids exposing the molecular layer to vacuum, but it also mitigates physical damage to the SAM from the direct application of solid metallic contacts. 394Unlike nTP, LOFO does not depend on chemical bonding or low surface free-energy elastomers, but it is constrained by capillary forces and requires exposure to solvent. 396The surface supporting the molecular ensemble must be kept wet until the metal film establishes an electrical contact, which is particularly amenable for making MEJs comprising proteins (see Sec. III D). 81 Au is often used in LOFO rather than Ag, Al, or Cu because detaching agents, such as hydrofluoric or acetic acid solutions, can etch or react with other metals, creating ill-defined interfaces between the top electrode and the molecular layer.To mitigate this limitation, Ikram et al. 397 reported a method for forming Ag and Al electrical contacts without using chemical etching treatment.Further modifications have led to related methods such as polymer-assisted LOFO (PALO), 386,398 wedge transfer, 399 and direct metal transfer. 400The PALO method, developed by Shimizu et al., 386 combines the advantages of LOFO and nTP, readily forming crossbar junctions, successfully overcoming wrinkling problems, and providing a potential path forward for integrated circuits of MEJs.

Nanopore junctions
The top electrode definitionally determines the area of an MEJ because it is installed last and, as in the aforementioned strategies, is smaller (in area) than the molecular layers.Shrinking the area of the   top contact is useful both to mitigate the influence of defects and to miniaturize devices comprising MEJs.While CP-AFM and crossed wires form nanoscopic contacts, they are not static, and neither LOFO nor nTP can reliably produce contacts that small.2][403] In general, the fabrication of a nanopore device starts with the creation of a bowl-shaped hole ($50 nm in diameter) on a Si 3 N 4 membrane using electron beam lithography and reactive ion etching [Fig.21(a)]. 402The bottom electrode ($200 nm Au) is thermally evaporated onto the top side of the membrane, filling the pore.
A SAM is grown on the bottom electrode, and then a top electrode is evaporated such that the total area of the MEJ is defined by the diameter of the hole in the Si 3 N 4 .Alternatively, a layer of SiO 2 is evaporated on a bottom electrode and a focused ion beam is used to drill a nanowell ($40 nm in diameter) through to the bottom electrode upon which a SAM is grown before a top electrode is deposited to complete an MEJ that is defined by the diameter of the hole etched in the SiO 2 403 [Fig.21(b)].The essential feature of nanopore junctions is the area of the MEJ, which incorporates a small number of self-assembled molecules ($1000), comparable to CP-AFM, minimizing the influence of defects and disorders that are present in larger junctions.Both nanopore methods also allow the fabrication of a large number of devices for potential integration and the collection of a statistically significant amount of data. 402Unlike CP-AFM, where the operator determines the exact placement of a junction on a large substrate, nanopore junctions exhibit unavoidable sample-to-sample variation in the electrical behaviors of junctions as a result of the multistep fabrication process.Additionally, the yields of the nanopore and nanowell devices can be very low (% 2%) due to the penetration of metal atoms into the SAMs during the deposition of the top electrode, although the pores are small enough that gaseous metal atoms tend to be slowed by collisions with the walls and yields can be improved by careful deposition.
Another approach to mitigating damage to the SAMs and increasing yields is to insert a conductive buffer layer between the top electrode and the SAM, which is discussed further in Sec.IV C.

Hanging mercury drops
The main disadvantage of solid metal top contacts is that they either have to be physically transferred to a molecular layer or vacuum deposited from the vapor phase.An alternative is to use metals that are liquids at room temperature, which can be physically transferred to make conformal contact to the molecular layer without vacuum deposition, resulting in the easy fabrication of high yields of MEJs. 404hough not in the modern context of ME, the use of liquid metal top contacts for molecular layers dates back to 1939 when Race and Reynolds 2 used liquid Hg to measure the electrical properties of "multimolecular films."They tried several metals, finding that Hg was unique in not damaging the molecular layer(s).In 1971, Mann and Kuhn 1 reported the measurement of tunneling currents through fatty acid monolayers using liquid Hg applied to Langmuir-Blodgett films supported by Al bottom electrodes.Becucci et al. 405 reported the use of Hg electrodes to mimic a biological membrane.Slowinski and colleagues 103,406,407 fabricated Hg/SAM//SAM/Hg and Hg/SAM/Hg junctions (where '/' and '//' represent covalent and van der Waals interfaces, respectively) by bringing two mercury drops, one or both of which bearing a SAM, in contact using a micromanipulator and applying a bias across the drops.This approach exploits the ability to grow defect-free SAMs on the surface of Hg.Simply exposing a Hg drop to a solution of alkanethiols is enough to form a pinhole-free SAM, enabling the extensive study of charge tunneling in symmetric and asymmetric alkanethiol junctions as well as the relationship between structure and stability. 104,408Asymmetric junctions can be formed by bringing a hanging mercury drop electrode (i.e., a drop of Hg hanging from a conductive lead) into contact with a conductive substrate supporting a monolayer (as described in Sec.II).One limitation of using Hg is the formation of amalgams with Au and Ag; 19,409 a single pinhole defect can precipitate the dissolution of the entire bottom electrode.This problem is somewhat mitigated by the passivation of the Hg top electrode with an additional SAM of alkanethiols-pinhole free because it is formed on Hg-and measuring charge transport by contacting it with SAM-modified 410,411 or bare-bottom electrodes, 186,412 e.g., an oxide-free silicon substrate as shown in Fig. 22.These methods permit measurements of a large number of tunneling junctions on the same or different substrates, which then facilitate the collection of large sets of data for statistical analysis and interrogating reproducibility.Unfortunately, Hg has a non-zero vapor pressure at room temperature and is toxic, and although it can be handled safely in a laboratory, it obviates commercialization.It is also difficult to encapsulate and otherwise requires measurements to be performed in solvent baths, limiting even laboratory-scale studies on protodevices.

Capillary tunneling junctions
Fan et al. 414 and Liu et al. 415 fabricated MEJs inside capillary tubes in which the molecules self-assemble on Sn or In electrodes to form a capillary tunnel junction (CPT).These junctions are formed inside capillary fibers between solidified tin/indium electrodes, as illustrated in Fig. 23.Both Sn and In are low-melting metals that form native oxides and are well known in optoelectronic devices.To form a CPT, melted Sn or In is allowed to infiltrate a capillary fiber to a preferred height at temperatures that are slightly higher than their melting points.The metal is then cooled until it solidifies.Second, the end of the fiber is exposed to a solution of alkanethiols or carboxylic acids to form a SAM at the metal/metal-oxide surface.Finally, the SAMmodified fiber is inserted into a larger fiber that is also filled with melted Sn or In, encapsulating the SAM between two electrodes.The MEJ is completed when the junction is cooled to room temperature.These MEJs are both stable and reproducible.Li et al. 416 also reported the use of Ga in CPTs to investigate tunneling conductance of arylhalides.The primary limitation of this method is that the contact area and the distance between the two electrodes are fixed when the MEJ is formed; liquid metal electrodes that remain liquids during measurement can accommodate changes in the molecular layer, such as photo-induced isomerization. 417Another disadvantage of this method is that the insertion of the smaller fiber into the larger one may cause mechanical damage to the molecular layer, limiting the kinds of SAMs that can be used.

Eutectic gallium-indium
The liquid metal alloy EGaIn was first proposed by Chiechi et al. 56 as a nontoxic alternative to hanging mercury drop electrodes.Although it is a liquid metal, it shares few properties with Hg.It does not form amalgams and is easily encapsulated, and measurements can be performed in ambient conditions.Moreover, it can be molded into cone-shaped tips (Fig. 24) and used to form nondamaging contacts to SAMs that are % 10 times smaller than those formed using Hg.The reversible, nondamaging nature of EGaIn also allows extensive studies in the correlation between molecular structure and electric properties in MEJs by combining electrical measurements with surface spectroscopies in situ and ex situ.It is also useful as an electrode for impedance spectroscopy on soft, fragile films. 267,418  The non-Newtonian rheology of EGaIn allows it to form stable microstructures upon injection into microfluidic channels. 57,420ijhuis et al. 421 described a method of fabrication that generates crossbar arrays of MEJs based on SAMs.These junctions stabilize EGaIn top contacts in microchannels and use ultraflat (template-stripped) bottom electrodes to achieve yields of 70% to 90%.The technique rapidly generated large sets of data (N ¼ 300-800) and allowed measurements over a broad range of temperatures (100 K to 393 K), allowing variable-temperature charge-transport experiments.They further compared the results of this method with those from conical EGaIn tips, concluding that both produce high device yields and indistinguishable results. 422However, this method requires patterned bottom electrodes and the resulting, unavoidable edge effects and contamination by photoresist can influence the junction properties.Wan et al. 423 solved those problems by combining filling through-holes in PDMS with EGaIn and bringing them into contact with SAMs.This micropore method does not need patterned bottom electrodes and is compatible with large, ultraflat, template-stripped surfaces.Similar to the aforementioned photocurable polymer scaffolds, it immobilizes PDMS in a polymer form but in a barrel shape rather than a tip.A single top electrode can be used to form up to 15-25 junctions and yields highly reproducible electrical characteristics.Karuppannan et al. 424 used AlO x micropores instead of PDMS to form EGaIn MEJs on templatestripped Au surfaces, using a double-etch process to avoid contamination and unwanted etching of the Au surface and minimizing stray capacitance and leakage currents in the junction.This method enables variable-temperature measurements down to 8.5 K, has excellent current retention characteristics, and forms devices that are stabile for at least 2 months.

C. Hybrid top electrodes
Here, hybrid top electrodes refer to the use of a buffer layer between the top electrode and the molecular layer that prevents the penetration of metal atoms into the layer, reducing or eliminating electrical shorts and enabling vapor deposition and many common photolithographic techniques.These advantages facilitate the fabrication of robust and stable devices that can be integrated into standard semiconductor fabrication lines.Common materials for buffer layers include nanoparticles, [425][426][427] aluminum oxide, [428][429][430] conductive polymers, 33,431,432 graphene or graphene oxide sheets, [433][434][435] and carbon paint. 436an Hal et al. 437 first reported a method for fabricating highyielding MEJs that are compatible with semiconductor fab techniques that incorporates the conductive polymer blend poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) as a buffer layer between the top electrode and the SAM. 33These micropore MEJs are defined by microscale vertical interconnections-via holes photolithographically patterned in a photoresist-that eliminate parasitic currents in crossbar arrays and completely encapsulate the thiol-SAMs within.The yields of working devices are % 100% in devices that are air stable for at least several months without any signs of degradation, yielding identical I/V sweeps month after month that scale perfectly with the diameter of the via holes, which can be as big as 100 lm.Van Hal et al. 437 demonstrated compatibility with semiconductor fabrication technology and the ability to upscale micropore MEJs integrated with silicon semiconductor devices.However, the use of PEDOT:PSS imposes several limitations, such as a limited processing window (<50 C), a poorly defined polymer-SAM interface, and a strong dependence of yields on the type of molecules used in the SAMs. 438euhausen et al. 432 reported the use of Aedotron P-a low-viscosity, amphiphilic conducting polymer-as a replacement for PEDOT:PSS that allows the incorporation of SAMs with either hydrophobic or hydrophilic end groups without changing the junction architecture.The main limitation of micropore MEJs is the contact resistance of polymer buffer layers; PEDOT:PSS and its derivatives and Aedotron P all yield contact resistance that will dominate the electrical properties of junctions comprising all but the most resistive SAMs, e.g., alkanethiols.The all-carbon MEJs discussed in this section and in Sec.II B 2 capture most of the useful properties of micropore MEJs without the contact resistance issues, although with a different set of trade-offs vis-a-vis upscaling, integration, and tolerance for functional groups.
A nanometer-thick aluminum oxide (Al 2 O 3 ) layer can also be used as a protective layer on SAMs to avoid damages to the molecules and shorts in the circuit caused by direct metal deposition. 428Atomic layer deposition (ALD) processes in which the SAMs are exposed to the precursors trimethyl aluminum (TMA) and water vapor create an Al 2 O 3 monolayer in each cycle.Non-shorting devices can be fabricated with the thickness of Al 2 O 3 of 1 nm to 4 nm.The Al 2 O 3 layer with thickness of 1.5 nm to 2 nm acts as an efficient tunneling barrier for spin injection, 439 which assists the study of spin-selective electron transport in SAMs comprising chiral conducting polymers. 430owever, the application of Al 2 O 3 -buffer layers is limited to molecules with hydrophilic terminal groups. 440In the context of laboratory-scale studies, EGaIn solves the contact resistance problem by relying on the subnanometer, self-limiting Ga 2 O 3 as an in-built buffer layer, 441 although (as described in Sec.IV B 3) devices incorporating EGaIn top electrodes are less technologically useful than micropore or all-carbon MEJs.
The thinnest buffer layer that is physically possible is one atom thick, i.e., a single sheet of graphene.Wang et al. 433 first reported the use of multilayer graphene as a buffer layer in MEJs.The graphene buffer layers were grown on Ni substrates by chemical vapor deposition (CVD) and physically transferred on top of SAMs in SiO 2 micropores.At the time, the authors were only able to produce % 10-nmthick graphene multilayers by CVD (i.e., not one-atom-thick sheets).Statistical analysis of the electrical properties of the MEJs showed that their graphene-based devices had very low contact resistances, comparable to the metal-molecule-metal devices without buffer layers and could be obtained in yields of % 90% regardless of the (hydrophobic or hydrophilic) SAMs used.Thus, buffer layers of multilayer graphene were shown to produce excellent device durability, thermal stability, and longer lifetimes compared to other MEJs without a buffer layer and lower contact resistance than MEJs with polymer buffer layers.Li et al. 434 and Seo et al. 442 nearly simultaneously demonstrated the use of reduced graphene oxide (rGO) to avoid the limitations of CVD.In this method, graphite is exfoliated by oxidization to form suspensions of graphene oxide that are solution proceeded into films that are then reduced (thermally or chemically) to rGO.These % 5-nm-thick layers are then transferred to SAMs in Al 2 O 3 micropores, and a protective layer of Au is installed on top by vapor deposition.Initially, rGO did not perform as well as multilayer graphene, affording higher contact resistance and introducing a defect-mediated thermally activated hopping component; however, K€ uhnel et al. 435 have since refined the technique to produce rGO-MEJs that operate from room temperature to <1 K.The contact resistance of these improved rGO buffer layers is similar to that of pure metal-molecule-metal junctions; however, installing the buffer layer still requires a LOFO step, which presents a challenge for upscaling and integration.
Polymer and graphene/rGO buffer layers constrain the type and magnitude of electrical functions that are possible.Puebla-Hellmann et al. 427 reported the use of Au nanoparticles (AuNPs) as the top electrode to retain the advantages of buffer layers without the constraints.The AuNPs are processed from solution directly onto SAMs, directly forming a metal-molecule interface that also serves as a conformal, protective layer for the installation of top electrodes by vapor deposition, as depicted in Fig. 25(a).This approach enables the fabrication of thousands of identical, room-stable metal-molecule-metal MEJs.The I/V characteristics and IETS spectra of devices comprising SAMs of alkanethiols of different lengths produced values of b commensurate with charge transport that is dominated by the molecular layer and not strongly affected by the AuNP or metal overlayer.However, this method also has an important limitation: The AuNP layer requires SAMs that present anchoring groups to adhere to, which limits the scope of molecules that can be studied.This requirement is shared with micropore MEJs and is known to complicate the self-assembly process, often leading to lying-down or looped SAM phases that can also result in multilayer formation. 443lectrobeam-deposited carbon (eC) 215 and carbon paint also have been developed as buffer layers to form stable MEJs.Najarian et al. 215 reported molecular junctions fabricated with eC between Copyright 2020 American Chemical Society. 436AMs and both metal electrodes.The buffer layer of eC between the Au bottom electrode and the molecular layer reduces the roughness of the substrate and prevents electromigration.The buffer layer of eC between the Au top electrode and the molecular layer imparts longterm device stability and eliminates electromigration and oxidation.Moreover, partial transparency of one or both contacts permit in operando monitoring of MEJs with optical spectroscopy as well as photocurrent electroluminescence phenomena.Karuppannan et al. 436 used protective layers based on commercially available carbon paint to form high-quality, large-area MEJs with low contact resistance as shown in Fig. 25(b).This protective layer has the advantage of being solution-processable, eliminating the need for LOFO, which is required by graphene and rGO.Carbon paint buffer layers also allow measurements over a wide range of temperatures (8.5 K to 340 K).The resulting MEJs are also stable and reproducible and can be fabricated with 100% yield of non-shorting junctions.The absolute resistance of carbon paint is higher than graphene and rGO (but comparable to eC), limiting it to MEJs comprising SAMs that are more resistive than butanethiol.

D. Nonmetallic top electrodes
Nanoscale metal electrodes are mechanically unstable due to the high mobility of the metal atoms.While MEJs comprising solid-metal electrodes remain operational, experimental platforms based on nonmetallic materials have been developed, thus leading to new possibilities for molecular-scale electronics.Carbon-based and silicon-based electrodes, in particular, have been developed to fabricate stable, highyielding molecular devices.Carbon-based nanomaterials, including the single-walled carbon nanotubes (SWCNTs), 444 graphene sheets, 236,433,434 rGO, 435,442,445 eC, 215 and carbon paint, 436 were developed as point contacts, which are promising for progress in fabrication of electronic devices.In Sec.IV C, we discussed graphene, rGO, eC, and carbon paint buffer layers to form a soft contact to the SAM and to protect it from subsequent vapor deposition of a solid metal top contact.
Here, we discuss the use of SWCNTs and graphene/rGO as the top electrode or both the electrodes in MEJ.SWCNTs and graphene have sp 2 -hybridized carbon atoms arranged in a honeycomb lattice, offering natural compatibility with organic/biological molecules. 446He et al. 444 reported a testbed containing two oxide-free, nonmetallic electrodes, silicon and SWCNTs.This metal-free testbed eliminates the issue of the metal filament formation in MEJs, avoiding fatigue-induced shorts.They observed hysteretic I/V curves for p-conjugated oligophenylenethynylene molecules in these silicon-molecule-SWCNT MEJs, which can be useful toward memory storage applications.Control devices without of p-conjugated molecules (bare Si-H or long-alkane molecules) produced no hysteresis, indicating that the observed memory effect was intrinsic to the molecules (in combination with these electrodes).
Thin graphene and reduced graphene oxide sheet are optically transparent, making them good candidates for fabricating photoswitching devices.Li et al. 445 reported the use of solution-processed, ultrathin rGO as a transparent, soft top contact to successfully fabricate solid-state switches operated by thermo-optical stimuli.Seo et al. 236 reported the fabrication of an MEJ between two graphene electrodes in which the two ends are chemically and physically bonded to the top and bottom graphene electrodes, respectively.This configuration imparts stability and improved yields due to the soft top contact.Moreover, the transparent graphene electrodes allow stable molecular switching induced by external light stimulus.
Silicon-based materials benefit from decades of intense development in the semiconductor industry and compatibility with complementary metal oxide semiconductor technology, such as controllable conductivity and the tuning of E f by doping. 447The work function of silicon depends on the doping, type, and density of charge carriers but not on the interfacial structure and the chemical binding, which allow for the independent interfacial energy alignment.This distinction is also the reason silicon materials are widely used as bottom electrodes, as is discussed in Sec.III.9][450][451] First, a vertical Si/SiO 2 /Si structure is formed in which the bottom and top electrodes are the heavily doped Si (111) and polycrystalline Si surfaces, respectively.The native oxide layer is then etched using NH 4 F to create a nanogap.The nanogap serves as a platform for the formation of thin films or assembly of nanoparticle arrays. 449Further, the nanogap can be scaled down to include only a few molecules or a single molecule in the junction between the two silicon electrodes.

V. FUTURE OUTLOOK AND CHALLENGES
For decades, research in molecular electronics was mostly fundamental; to the extent that it addressed technological concerns, it was a solution (ultra-miniaturization) waiting for a problem (the physical limits of conventional semiconductor materials).Studies on SMJs advanced the theoretical descriptions of how tunneling electrons interact with the electronic structure of individual molecules.Studies on MEJs focused on alkanes and the difficult problem of isolating molecular phenomena from artifacts caused by low yields, defects, etc., or demonstrating the utility of a fabrication method.In recent years, owing in part to improvements in top electrode materials and methods, MEJs have been used to demonstrate useful functionality and interesting phenomena that are often not possible in SMJs due to their transient nature or the lack of intermolecular effects-or, more significantly, are (nigh) impossible to realize with conventional semiconducting materials.While important advances continue with SMJs, this review focuses on MEJs because of this increasingly rapid maturation.To the extent that molecular electronics leads to useful technology, SMJs will be the limiting case, but MEJs are the pathfinders.A key to this progress is the steady improvement in stability, with Nheterocyclic carbene, non-covalent graphene ligand, and fullerene anchoring groups all retaining the useful properties of self-assembly while dramatically improving on the stability of thiol-SAMs, mitigating the necessity for rigorous encapsulation.Carbon monolayers and silane monolayers further increase this stability across the threshold for commercialization.These optimizations to the formation of molecular monolayers on bottom electrodes have been synchronous with advances in graphene, carbon, patterned liquid metal, and other top electrodes and an ever-increasing understanding of the details of the interplay between molecular orbitals and charges traversing MEJs.
Research in MEJs is unquestionably generating interesting scientific discovery and insight while demonstrably advancing technologically in three key areas.First, comparisons to silicon semiconductor technology are increasingly obviated by scientific advances.Such comparisons were once commonplace, but miniaturization is a solved problem, and molecular electronics will never be competitive with an established, 60-year-old multibillion-dollar industry.The focus is shifting to the unique function of MEJs, be it electrical output characteristics, the ability to leverage discrete intermolecular interactions, quantum interference phenomena, or something else.Second, junctions are approaching useful degrees of stability: months and years, not minutes and hours, and hundreds of thousands of cycles, not dozens.Excluding special cases like disposable sensors, electronic devices cannot die after a day of use.Third, electrical inputs and outputs that are compatible with silicon semiconductors, for example, current densities and switching ratios, are reported with increasing frequency.Just as quantum computing circuits are programmed and read through conventional, digital circuits, any application of MEJs will require them to interface with conventional electronics.While there are obviously many engineering problems that will need to be solved on the road to technologies based on MEJs, these three scientific problems already have plausible and increasingly robust solutions, some of which have emerged only in the past few years.
Molecular electronics has a bright future, with many laboratories now specializing in either MEJs or SMJs.With more sophisticated theory and growing computational resources, the virtuous cycle between experiment and theory continues to bring more researchers into the field to produce ever more impressive results.It is evident, even from the relatively brief treatment of bottom electrodes (Sec.II), molecular layers (Sec.III), and top electrodes (Sec.IV) in this review, that the essential components and proofs of concept necessary to solve important, emerging technological problems and foment breakthroughs exist and that the challenge in MEJs is to combine them to affect these goals.

AUTHORS' CONTRIBUTIONS
X.Q. and S.S. contributed equally to this work.
FIG.1.A conceptual summary of molecular ensemble junctions.An ensemble of molecules, often a single monolayer, is sandwiched between two electrodes (gray and gold spheres) to form a junction that can be completely symmetric, utilize electrodes of different materials, and/or comprise asymmetric molecules.The constituent molecules can be alkanes, shown here with two common types of anchoring groups, p-conjugated molecules, organometallics, or biomolecules.In each case, the molecular properties translate into useful electrical outputs, such as switching and rectification.

FIG. 2 .
Plots of typical data acquired from a molecular ensemble junction.(a) 230 raw J/V traces (blue lines), mean values (black circles), and the standard deviation (black error bars).(b) The histogram of log jJj at V ¼ 1:0V (blue bars) and Gaussian fit (pink line) of the raw J/V data.The dashed box in (a) indicates the data plotted in the histogram in (b).The black circle and error bar in (b) approximate how the data in (a) are plotted from the Gaussian fit.The data marked "short" were acquired from the junction that failed, i.e., an electrical short.

FIG. 3 .
FIG. 3. Different statistical methods (de)emphasize features of the underlying dataset that can affect their physical interpretation.(a) Deviations of log jJj from normality and their effects on methods 1-3 discussed in the text.(b) Schematic of different statistical methods of analyzing charge transport discussed in Sec.I C. Reprinted with permission from Reus et al., J. Phys.Chem.C. 116, 6714-6733 (2012).Copyright 2012 American Chemical Society.58

FIG. 4 .
FIG. 4. Log-normal distributed conductance data contain information that can be revealed by the method of visualization.(a) Discrete data of unhealthy measurement of tunneling junctions; histograms of J at different values of V in terms of counts (top) and the corresponding heat map plot (bottom).Reprinted with permission from Sporrer et al., J. Phys.Chem.Lett.6, 4952-4958 (2015).Copyright 2015 American Chemical Society. 59(b) Schematics of a Gaussian fit to a histogram of J and a description of first and second statistical moments: Gaussian mean and standard deviation; the third and the fourth statistical moment (skewness and kurtosis).Reprinted with permission from Chen et al., J. Phys.Chem.Lett.9, 5078-5085 (2018).Copyright 2018 American Chemical Society.60

FIG. 5 .
FIG. 5.A mixed monolayer of thiols assimilates bulky fullerene cages.(a) Schematic diagram of a junction prepared by incubating SAMs of short alkanethiols in a solution of fullerenes bearing alkylthiol linkers.(b) Plots of log jJj vs V before and after incubation with the fullerene derivative.The J/V curve is symmetric before the incorporation of the fullerene, which alters the charge-transport mechanism at positive bias, inducing rectification.Reprinted with permission from Qiu et al., Chem.Sci. 8, 2365-2372 (2017).Licensed under a Creative Commons Attribution (CC BY) license.25

FIG. 7 .
FIG. 7.An MEJ comprising electrochemically grafted molecules on carbon electrodes.(a) Schematic of the junction in which a molecular bilayer of fluorene (FL) and Li þ -bound benzoic acid (LiBA) is electrochemically grafted onto the surface of electron beam-deposited carbon electrode supported by Au.Reprinted with permission from Qiu et al., Trends Chem.2, 869-872 (2020).Copyright 2020 Elsevier. 218(b) Optical image of four junctions fabricated on a SiO 2 /Si substrate.(c) J/V response of the junctions in vacuum (vac; red curve), of the initial scan in air (blue curve), and after several scans in air (pink curve).Reprinted with permission from Mondal et al., J. Am.Chem.Soc.140, 7239-7247 (2018).Copyright 2018 American Chemical Society.214 incorporated oligophenylenes into MEJs comprising NHC-SAMs to study their tunneling charge-transport and thermoelectric properties [Fig.9(c)].That work also demonstrated the formation of densely packed SAMs on ultraflat, polycrystalline Au substrates instead of the Au(111) previously used to study NHC-SAMs, which reproduce values of b reported for thiol-SAMs [Fig.9(d)].

FIG. 8 .
FIG.8.MEJs comprising SAMs of carboxylates, alkynes, and thiolates show indistinguishable charge-transport properties.(a) Plot of the Gaussian mean values of log jJj at þ0.5 V vs molecular length (calculated in A ˚for an all-trans-extended conformation) for alkyl SAMs with A ¼ -S-, -CC-, and -CO2-anchoring groups; the distance is calculated from the anchoring atom A that covalently binds to the surface of metal substrates to the distal hydrogen atom.(b) Plot of the Gaussian mean values of log jJj at þ0.5 V vs number of methylene units.Reprinted with permission from Bowers et al., ACS Nano 9, 1471-1477 (2015).Copyright 2015 American Chemical Society.223

FIG. 9 .FIG. 10 .
FIG. 9. Reactions of NHCs on sulfide/NHC-protected Au surfaces and characterizations on the electrical properties of NHC-SAMs.(a) XPS evidence for the complete loss of dodecyl sulfide on treatment with NHCs as demonstrated by the lack of an S (2p) signal.(b) XPS analysis of the treatment of NHC-protected Au surfaces with dodecanethiol shows no incorporation of sulfur.Reprinted with permission from Crudden et al., Nat.Chem.6, 409-414 (2014).Copyright 2014 Springer Nature. 229(c) Schematic of an MEJ comprising NHC-SAMs of oligophenylenes and EGaIn top electrode.(d) Linear correlation between current density and molecular length in semi-log scale on NHC-SAMs of oligophenylenes.(e) Plots of thermoelectric voltages created by the temperature differential applied to the junctions comprising NHC-SAMs of oligophenylenes.Reprinted with permission from Kang et al., Chem.Commun.55, 8780-8783 (2019).Copyright 2019 Royal Society of Chemistry.235 FIG. 11.The SAMs and SABs of PTEG-1.(a) Schematic of the transition from a SAM of PTEG-1 into a bilayer.[(b)-(d)] AFM height maps showing the growth stages of bilayers of PTEG-1 on Au substrates: (b) the monolayer, (c) the incomplete bilayer, and (d) the complete bilayer.(e) Plots of log jJj vs potential for MEJs comprising SAMs and SABs of PTEG-1 and EGaIn top electrodes.'/' denotes interfaces defined by chemisorption and '//' by physisorption.[(g) and (h)] STM images of a SAM of PTEG-1 on Au(111)/mica showing the presence of C 6 O cages under a tip bias of (g) 0.3 V and (h) -0.3 V, and with a set point of 150 pA.The inset in (h) is a fast Fourier transform of the STM image giving information on the arrangement of PTEG-1 in the monolayer.Reprinted with permission from Qiu et al., Nat.Mater.19, 330-337 (2020).Copyright 2020 Springer Nature.170

FIG. 12 .
FIG. 12. Schematic illustrations showing MEJs comprising SAMs of four different classes of molecules between two metal electrodes, where R, X, and Y are functional groups.(a) Aliphatic molecules with functionalized and saturated backbones (Sec.III A); (b) r-p molecule schematic containing saturated tail and a conjugated head group (Sec.III B); (c) fully p-conjugated molecules (Sec.III C); and (d) biomolecules, in this case a photosystem I complex (Sec.III D).

2 FIG. 13 .
FIG. 13.Examples of several series of aliphatic molecules discussed in Sec.III A. (a) Alkanethiolate SAMs with varying lengths showing exponential decay of current with increasing molecular length.Reprinted with permission from Bowers et al., Nano Lett.14, 3521-3526 (2014).Copyright 2014 American Chemical Society. 257(b) Series of alkanethiolate chains with the same length but varying head groups functionalized with different functional groups at the SAM//EGaIn interface.Reprinted with permission from Baghbanzadeh et al., ACS Nano, 12, 10221-10230 (2018).Copyright 2018 American Chemical Society. 225(c) The odd-even effect manifesting in a series of aliphatic chains with an increasing number of carbon atoms in the chain.Reprinted with permission from Thuo et al., J. Am.Chem.Soc.133, 2962-2975 (2011).Copyright 2011 American Chemical Society. 35(d) Series of alkanethiolates with different halogen atoms as head groups, exhibiting a relationship between the current density and dielectric constant of SAMs as a function of the polarizability of the halogen atom.Reprinted with permission from Wang et al., Adv.Mater.27, 6689-6695 (2015).Copyright 2015 John Wiley and Sons. 267(e) Aliphatic chains of the length with amide groups in the core that induce dipole moments with varying orientation, causing shifts in asymmetry in tunneling current behavior.Reprinted with permission from Baghbanzadeh et al., J. Am.Chem.Soc.141, 8689-8980 (2019).Copyright 2019 American Chemical Society.265WF, work function.

FIG. 14 .
FIG. 14.Molecular junctions comprising SAMs of molecules with r-p frameworks.(a) The length dependence of r-p systems containing oligothiophenes connected to butylthiol tails that exhibit a nonlinear dependence, i.e., increasing conductance with chain length past T2C4.Reprinted with permission from Zhang et al., J. Am.Chem.Soc.140, 15048-15055 (2018).Copyright 2018 American Chemical Society. 95(b) The reversal of the polarity of rectification by varying the position of a ferrocene moiety along the length of aliphatic chain.Reprinted with permission from Yuan et al., Nat.Commun.6, 6324 (2015).Copyright 2015 Springer Nature. 250(c) Methyl-viologen-terminated molecules showing double functionality ascribed to counterion migration across the MEJ.Reprinted with permission from Han et al., Nat.Mater.19, 843-848 (2020).Copyright 2020 Springer Nature. 99(d) Molecular rectifiers on silicon bottom substrate, enhancing the performance by varying the p head group, changing the interface coupling with the top electrode.Reprinted with permission from Lamport et al., ACS Appl.Mater.Interfaces 11, 18564-18570 (2019).Copyright 2019 American Chemical Society.281

FIG. 15 .
FIG. 15.Examples of molecular junctions comprising SAMs of fully conjugated molecules governing tunneling charge-transport characteristics.(a) Two-terminal memory storage protodevices comprising of SAMs of tetracyano-substituted anthraquinoid cores that can undergo reversible redox with the bottom substrate upon application of bias.Reprinted with permission from Carlotti et al., Angew.Chem.Int.Edit.57, 15681-15685 (2018).Licensed under a Creative Commons Attribution (CC BY) license. 297(b) Vertical transistors based on destructive quantum interference in fully conjugated molecular wires.Reprinted with permission from Famili et al., Chem 5, 474-484 (2019).Copyright 2019 Elsevier. 295(c) Oligophenylene mono-and dithiol derivatives of varying lengths studied in MEJs by CP-AFM and analyzed using a single-level model to understand the effect of functional groups on the coupling strength at the molecule-electrode interface.Reprinted with permission from Xie et al., J. Am.Chem.Soc.141, 3670-3681 (2019).Copyright 2019 American Chemical Society. 301(d) The effect of dipoles at SAM//EGaIn interfaces on the work function, transition voltage, and current density exhibited by Asyuda et al. 302 in binary SAMs comprising biphenyl SAMs terminating with CF 3 and CH 3 in different proportions.Reprinted with permission from Asyuda et al., J. Phys.Chem.C. 124, 24837-24848 (2020).Copyright 2020 American Chemical Society.302

FIG. 16 .
FIG. 16.Biomolecule-based junctions.(a) Devices utilizing Az, bacteriorhodopsin (bR), and bovine serum albumin (BSA) as monolayers retaining their native conformation facilitating electric current follow in the solid state better than a control saturate organic molecule (OTMS).Reprinted with permission from Ron et al., J. Am.Chem.Soc.132, 4131-4140 (2010).Copyright 2010 American Chemical Society. 81(b) A solid-state protein MEJ comprising Az between Au nanowire and bottom substrate, showing different I/ V behavior without (left) and with (right) mercaptopropionic acid as a linker.Reprinted with permission from Fereiro et al., Proc.Natl.Acad.Sci.U. S. A. 115, E477-E4583 (2018).Licensed under a Creative Commons Attribution (CC BY) license. 334(c) The self-assembly of photosystem I in MEJs where the current behavior is shown to be dependent on the orientation of the protein complexes.Reprinted with permission from Castañeda Ocampo et al., J. Am.Chem.Soc.137, 8419-8427 (2015).Copyright 2015 American Chemical Society. 80(d) An electrochemical STM setup to measure charge transport through PSI complexes on SAM of 8-mercaptooctanoic acid (MOA) on Au surfaces, showing the cyclic voltammogram I/V sweeps and photocurrent on the left and STM surface scans of PSI on Au on the right.Reprinted with permission from L opez-Mart ınez et al., Angew.Chem.Int.Edit.58, 13280-13284 (2019).Copyright 2019 John Wiley and Sons.335

FIG. 17 .
FIG. 17.Two different strategies usually implemented for fabricating mixed monolayers.(a) This strategy includes one shorter diluent molecule, e.g., an alkanethiol or an oligo(ethylene glycol), and a second molecule with a bulky head group and long, thin flexible linker.A few examples of bulky head groups attached to a linker are spiropyrans, azobenzene, C60, bipyridyl, and ferrocene.(b) The second strategy involves incorporating two molecules "A" and "B" with similar backbone structure but with different functionalization, different anchoring groups, different orientation on the surface.Some examples include biphenylthiols with different termini, alkanethiols with different lengths, aliphatic molecules with disulfide or thiol anchors, and differently oriented PSI protein.

FIG. 18 .
FIG. 18. Schematic description of a crossed wire junction.The contact between the wires is controlled by the Lorentz force.One wire is coated with a SAM perpendicular to the applied magnetic field B. The low current through the wire coated with a SAM generates a Lorentz force, and the two wires are then gently brought together to form a junction at the contact point.Inset: I/V characteristics of metalmolecule-metal junction.Reprinted with permission from Kushmerick et al., Nano Lett.3, 897-900 (2003).Copyright 2003 American Chemical Society.376

FIG. 19 .
FIG. 19.The nanotransfer printing procedure.(a) A GaAs substrate is first etched in concentrated NH 4 OH or HCl and then immediately exposed to a 1,8-octanedithiol vapor or solution for self-assembly.(b) The gold-coated elastomeric PDMS stamp is brought into contact with the modified substrate.(c) The stamp is removed from the substrate, and Au on the PDMS stamp is bonded by the molecules and transferred to the molecule-coated GaAs substrate.Reprinted with permission from Loo et al., Nano Lett.3, 913-917 (2003).Copyright 2003 American Chemical Society.384

FIG. 21 .
FIG. 21.The fabrication of nanopore devices.(a) Nanopore junction.A pore is made through a Si 3 N 4 membrane via reactive ion etching.The bottom Au electrode is evaporated onto the other side of the membrane, upon which a SAM is formed before completing the junction with the evaporation of the top electrode.(b) Nanowell junction.A hole is drilled through a layer of SiO 2 with a focused ion beam layer to expose a bottom electrode of Au.After growing a SAM on the electrode, the junction is completed with the deposition of an Au top electrode.Reprinted with permission from Xiang et al., Chem.Rev.116, 4318-4440 (2016).Copyright 2016 American Chemical Society.21

FIG. 23 .
FIG. 23.Capillary tunneling junctions.(a) Molten metal (Sn or In infiltrates a capillary fiber due to capillary forces).(b) After solidifying, the exposed end of the fiber is then immersed in an ethanolic solution of the target molecules to form a SAM.(c)The SAM-modified fiber is inserted into another fiber with a larger diameter also filled with molten metal, encapsulating the SAM.(d) A schematic of the CPT connected to an outer circuit.Reprinted with permission from Liu et al.,Langmuir 20,  855-861 (2004).Copyright 2004 American Chemical Society.415

FIG. 24 .
FIG.24.Optical micrographs showing the formation of a conical EGaIn tip.A microliter EGaIn droplet is squeezed out of a syringe and crashed into a bare metal substrate (first frame).As the syringe is retracted, it sticks to the substrate and, as shown in consecutive frames, forms an hourglass shape that eventually breaks off to form a sharp conical tip, as shown in the last frame.

2 FIG. 25 .
FIG. 25.Junctions with Au nanoparticle top electrodes.(a) Schematic of the fabrication of MEJs using Au nanoparticles as a buffer layer: (1) Pores of diameter d are etched into a dielectric film on a Pt electrode, (2) a SAM is grown on the platinum electrode, (3) AuNPs are deposited onto the SAM from solution, and (4) a top electrode is installed by vapor deposition.Reprinted with permission from Puebla-Hellmann et al., Nature 559, 232-235 (2018).Copyright 2018 Springer Nature. 427(b) Schematic showing junctions fabricated using carbon paint as buffer layer between the SAM and Au top electrode.Reprinted with permission from Karuppannan et al., J. Am.Chem.Soc.142, 3513-3524 (2020).Copyright 2020 American Chemical Society.436 376 419on et al.419reported a method for the in situ encapsulation of EGaIn microelectrodes with a