Single-Molecule Conductance Behavior of Molecular Bundles

Controlling the orientation of complex molecules in molecular junctions is crucial to their development into functional devices. To date, this has been achieved through the use of multipodal compounds (i.e., containing more than two anchoring groups), resulting in the formation of tri/tetrapodal compounds. While such compounds have greatly improved orientation control, this comes at the cost of lower surface coverage. In this study, we examine an alternative approach for generating multimodal compounds by binding multiple independent molecular wires together through metal coordination to form a molecular bundle. This was achieved by coordinating iron(II) and cobalt(II) to 5,5′-bis(methylthio)-2,2′-bipyridine (L1) and (methylenebis(4,1-phenylene))bis(1-(5-(methylthio)pyridin-2-yl)methanimine) (L2) to give two monometallic complexes, Fe-1 and Co-1, and two bimetallic helicates, Fe-2 and Co-2. Using XPS, all of the complexes were shown to bind to a gold surface in a fac fashion through three thiomethyl groups. Using single-molecule conductance and DFT calculations, each of the ligands was shown to conduct as an independent wire with no impact from the rest of the complex. These results suggest that this is a useful approach for controlling the geometry of junction formation without altering the conductance behavior of the individual molecular wires.


■ INTRODUCTION
−5 This has prompted the investigation of increasingly complex molecules and molecular assemblies as a way of achieving new electrical functionality of molecular junctions.However, as molecular junction complexity increases, the defined orientation and anchoring of molecular assemblies can become challenging.Here, the orientation of the molecule with respect to the substrate becomes critical in achieving defined junction morphology that maps onto a consistent and reproducible electrical response.
Current attempts toward improving molecular junction definition involve producing multipodal compounds with more than two contact groups or using contacts with very specific binding geometries.−11 In both cases, improved junction formations have been observed due to increased interaction with the electrode.However, the addition of multiple contact groups results in a large footprint, resulting in a lower surface coverage.Lower surface coverage may not be desirable since high surface coverage can promote better packing and ordering in self-assembled monolayers (SAMs) and this might better protect large-area molecular devices from short-circuiting as a result of the more stable and contiguous molecular monolayers.Therefore, several considerations such as stable and defined surface anchoring, reproducible electrical response, and surface assembly and stability are all important.In achieving these attributes, while developing new electrical functions, it is worthwhile exploring new methods to tether and assemble molecular wires within electrical junctions.
An alternative approach to assembling and tethering molecular junctions is to bundle multiple simple conductors together, thereby promoting a higher conductor density through geometric control.Shen et al. recently demonstrated that it is possible to join two parallel p-quaterphenyl molecular wires, which resulted in a slight increase in conductance (attributed to a through-space contribution). 12This study seeks to expand on Shen's work by combining three conductors into a bundle in a modular fashion; metallosupramolecular chemistry provides a convenient route for doing this.If each of the ligands used to build the complex structure is conjugated, then the addition of anchor groups produces molecular wires.One of the simplest of these is a 5,5′-substituted-2,2′-bipyridine.Although no conductance studies have been reported on 5,5′-substituted-2,2′-bipyridine complexes, previous work by Ponce et al. demonstrated that 3,8-substituted 1,10-phenanthroline (structurally similar to 5,5′-substituted-2,2′-bipyridine) showed little difference in conductance behavior when coordinated to a transition metal, which suggests that in this motif, the primary conductance path does not include the metal. 13However, in the case of [2,2′-bipyridine]-4,4′-diamine-based complexes, where conductance occurs through the metal−pyridine bond, the metal ion dictates the conductive behavior. 14In the work reported in this manuscript, iron(II) and cobalt(II) complexes were synthesized as both metal ions are labile and octahedral, giving isostructural complexes.When conductance occurs through a pyridine−cobalt(II) path, a significantly higher conductance is achieved relative to the analogous iron(II); 15 this observation provides a suitable evaluation of whether charge flow through the molecule is occurring through the metal center of the bundle.
As a means of testing the limits of junction stability, analogous complexes were examined with a high aspect ratio; therefore, a bimetallic metallohelicate was employed, based on the work of Hannon.The iron(II) (Fe-2) and cobalt(II) (Co-2) helicates were produced by first lithiating 2-bromo-5-(methylthio)pyridine, followed by treatment with DMF, to give the corresponding 5-(methylthio)picolinaldehyde.This was reacted with 4,4′-methylenedianiline to form the imine (L 2 ) in situ, 16 followed by the addition of a metal salt in a single-pot reaction.These reactions proceeded rapidly, with purification achieved by crystallization to give yields of Fe-2 (81%) and Co-2 (68%).It is noteworthy that although L 2 was formed in the absence of the metal salts, it could not be satisfactorily purified as the free ligand.Molecular Structures.Crystal structures of complexes Fe-1 and Co-1 showed typical tribipyridine complexes with a slightly distorted octahedral coordination environment of the metal centers (Figure 1), with Co-1 showing a marginally higher distortion of octahedral geometry around the metal atom (minimum N−Co−N angle 76.62 vs 81.82°in Fe-1, and average M−N bond lengths of 2.130 Å Co-1 vs 1.916 Å Fe-1) than in Fe-1.This was attributed to Co-1 being in the highspin (HS) state, whereas Fe-1 was in the low-spin (LS) state. 17his difference in geometries resulted in a more compact shape of cation Fe-1 in comparison with Co-1, which results in a closer spatial arrangement of S atoms in Fe-1: the range of S••• S distances there varied from 6.349 to 6.618 Å, while in the Co-1 complex, all of these distances were longer than 7.2 Å. Iron(II)-based helicate complexes have been described before, 18 and the geometry of the cation Fe-2 corresponded well to previously observed ones.As observed from Fe-1 and Co-1, the average M−N bond lengths were 1.979 Å (Fe-2) and 2.137 Å (Co-2), which was consistent with the former being LS and the latter being HS.The distance between the S 3 planes at the opposite sides of the cation was 19.435(2) Å.The S•••S distances in Co-2 (6.349−6.720Å) were shorter than in Co-1, probably due to the higher flexibility of the ligands.That led to a similar columnar packing arrangement to that found in Fe-2, with the distance between the opposite S 3 -planes in Co-2 being 20.065(2) Å (Figure 2).
XPS Data/Surface Coverage Data.In order to gain insights into the binding interactions between complexes and a gold substrate via the various potential thiol contacting groups, a surface study was performed with Fe-1 and Fe-2 using a quartz crystal microbalance (QCM) and X-ray photoelectron spectroscopy (XPS); see the Supporting Information for full details.This study will allow better understanding of the orientation of these complexes in the molecular junction (single-molecule conductance properties and DFT studies) through the use of multipodal contacts.It is noteworthy that only Fe-1 and Fe-2 were used for this study since Co-1 and Co-2 complexes show practically the same molecular structure, and there is not a direct participation of the metal centers in the interaction with a gold substrate.Additionally, Fe(II) is the most electrochemically stable of the metals and, therefore, would provide the most reliable results.
To monitor the SAM formation from Fe-1 and Fe-2, a QCM resonator was incubated in a 10 −4 M solution in acetonitrile for each compound, and its frequency variation followed with respect to the incubation time.After 24 h, no further frequency variation was observed for either compound.This method gave a surface coverage of the resulting selfassembly monolayers as 1.1 × 10 −10 and 1.2 × 10 −10 mol•cm −2 for Fe-1 and Fe-2, respectively, from the observed frequency variation of −14 Hz (Fe-1) and −31 Hz (Fe-2) and using the relationships described by the Sauerbrey equation. 19This shows that practically the same surface coverage (the same number of molecules per surface area) was obtained for both compounds, which suggests that Fe-1 and Fe-2 have the same orientation and arrangement in the SAM, even when Fe-2 is much longer.This result suggests that the complexes assemble in a vertical orientation with respect to the gold substrate interacting through some of the six potential thiomethyls, giving the same effective area per molecule.To obtain more details about the nature of the interaction between Fe-1 and Fe-2 and a gold substrate and hence gain information about the molecular orientation, XPS measurements were carried out on each of these iron compounds for both a powdered sample and SAMs on gold.
A powder sample of Fe-1 in the S 2p region (Figure 3) of the XPS spectrum displayed two peaks at 162.9 and 164.1 eV (with a peak separation of 1.2 eV and an area ratio of 2:1 (66 and 34%)), assigned to (2p 3/2 ) and (2p 1/2 ), respectively.As expected, this shows that all sulfur atoms in the powder sample are in a chemically equivalent environment.In contrast, a SAM of Fe-1 on a gold substrate gives a more complex XPS spectrum.The most intense pair of S 2p peaks at 163.0 and 164.2 eV is at practically the same binding energy as those observed for the powder sample.This indicates that at least some of the thiomethyls in Fe-1 are free; in other words, they do not interact with the gold substrate.On the other hand, the peak at the lower binding energy (161.6 eV) was attributed to a S 2p 3/2 peak, arising from thiomethyls interacting with the gold substrate. 7,20,21It would be expected that the corresponding weaker 2p 1/2 peak would fall ca.1.2 eV higher in energy (i.e., at 162.8 eV), but this is obscured by the more intense peaks from the thiomethyls that are not attached to the surface.
Therefore, the XPS data described above indicate that for a SAM of Fe-1, the molecules bind to the Au substrate through some of the potential thiomethyls contacting groups of Fe-1.
Given that each S 2p 3/2,1/2 doublet has a branching ratio of 2:1 (S 2p 3/2 /S 2p 1/2 ), and taking into account the relative intensities of the clearly observed peaks at 164.1 eV (nonbound thiomethyls, S 2p 3/2 ) and 161.4 eV (bound thiols, S 2p 1/2 ), the corresponding areas associated with each signal in the convoluted spectrum can be estimated at: 19.2% (164.2 eV), ∼38% (163.0 eV), ∼14.5% (∼162.8eV), and 29.4% (161.6 eV).From this, the relative area of XPS signals arising from thiomethyls not bound and bound to the gold substrate is approximated as 57:43.An estimate of the relative proportions of thiomethyls bound vs not bound was made, by using the attenuation of the Au 4f XPS signal (given the broadly similar binding energies of Au 4f and S 2p) from the gold substrate covered by a SAM of Fe-1 (Figure 3).An attenuation factor of 0.76 was estimated.Using this attenuation factor, if three out of the six potential thiomethyl contacting groups were not bound and three were bound in the film of Fe-1 on gold, this situation would be expected to give rise to XPS signals with an approximate ratio of 55:45.This ratio is very close to the integrated areas of the peaks observed in the experimental spectra (57:43).
The XPS data described above indicate that for the SAM of Fe-1, the molecules adsorb to the gold substrate through three of the six potential thiomethyl contacting groups; with the opposite three thiomethyls exposed on the top surface of the film and thereby available to contact a top electrode.In addition, the thickness of the SAM of Fe-1 was also obtained through an analysis of the attenuation of the Au 4f signal in the XPS spectra (Figure 3), using the relationship I SAM = I substrate exp(−d/λ sin θ), where the film thickness is d; I SAM and I substrate are the combined average of the intensities of the Au 4f 5/2 and Au 4f 7/2 peaks from the SAM and from the bare gold, respectively; θ is the photoelectron take-off angle (90°); and λ is the effective attenuation length of the photoelectron (4.2 ± 0.1 nm). 22From this analysis, a thickness of 1.14 nm is obtained, which is consistent with having the molecules adsorbed to the gold substrate through three of the six potential thiomethyl contacting groups, as observed by XPS.
A similar study was carried out for a SAM of Fe-2.In this instance, the XPS spectrum for a powder sample of Fe-2 in the S 2p region (Figure 3) showed two peaks at 163.4 and 164.6 eV (with a peak separation of 1.2 eV and an area ratio of 2:1 (63 and 37%)) assigned to (2p 3/2 ) and (2p 1/2 ), respectively.Meanwhile, a SAM of Fe-2 on a gold substrate provided an XPS spectrum with the most intense pair of S 2p peaks, at 163.5 and 164.7 eV, practically at the same binding energy as those observed for the powder sample.These are associated with the thiomethyls not interacting with the gold substrate.A peak at a lower binding energy (162.0 eV) is attributed to the S 2p 3/2 peak and arises from thiomethyls interacting with the gold substrate, 7,20,21 with the corresponding weaker 2p 1/2 peak being expected to fall ca.1.2 eV higher in energy (i.e., at 163.2 eV) but obscured by the more intense peaks from the thiomethyls not attached to the surface.As mentioned above, given that each S 2p 3/2,1/2 doublet has a branching ratio of 2:1 (S 2p 3/2 /S 2p 1/2 ), and taking into account the relative intensities of the clearly observed peaks at 164.7 eV (nonbound thiolmethyls, S 2p 3/2 ) and 162.0 eV (bound thiols, S 2p 1/2 ), the corresponding areas associated with each signal in the convoluted spectrum can be estimated at: 20.6% (164.7 eV), ∼41% (163.5 eV), ∼12% (∼163.2eV), and 24.2% (162.0 eV).That is, the relative area of XPS signals arising from thiomethyls not bound and bound to the gold substrate was approximately 64:36.Using the attenuation factor (0.56) of the Au 4f XPS signal from the gold substrate covered by a SAM of Fe-1 (Figure 3), an estimate of the relative proportions of thiomethyls bound vs not bonded was made.Again, if three out of the six potential thiomethyl contacting groups were not bound and three were bound in the SAM of Fe-2 on gold, this would give rise to XPS signals with an approximate ratio of 65:35 (i.e., very close to the integrated areas of the peaks observed in the experimental spectra).Therefore, these results indicate that for a SAM of Fe-2, the molecules adsorb to the gold substrate through three of the six potential thiomethyl contacting groups.The opposite three thiols are exposed on the top surface of the film and are thereby available to contact a top electrode, similar to compound Fe-1.Finally, using the attenuation of the Au 4f signal in the XPS spectra (Figure 3), a thickness of 2.4 nm was obtained, again consistent with SAM binding to the gold substrate through three of the six potential thiomethyl contacting groups.The XPS observations corroborate the QCM results; that a same surface coverage for Fe-1 and Fe-2 is indicative of a vertical orientation of the molecules with respect to the surface through three of the six potential thiomethyl groups.
Conductance.The conductances of complexes Fe-1 and Co-1 were measured using the STM-BJ technique recorded in propylene carbonate, with a molecular concentration of 1 mM and using an Apiezon wax insulated tip.The example traces and 2D histograms are given in the SI.All values for the singlemolecule conductance were obtained by fitting the corresponding 1D histograms with Gauss functions.The single-molecule conductance properties of the helical complexes (Fe-1 and Co-1) indicate that the metal centers do not actively participate in the dominant transport channel.This observation can be seen by comparing the results from both Fe-1 and Co-1; see Figure 4a,b.Here, the average conductance values are given by the peaks at 1 × 10 −2.77G/G 0 for Fe-1 and 1 × 10 −2.65 G/G 0 for Co-1.We measured the ligand L 1 by itself to compare how the conductance might change if no metal center is present.The average conductance value for the ligand is 10 −2.88 G/G 0 , which confirms that the contribution of the metals in the complexes is minimal, see Figure 4c.Since the magnitude of the conductance is similar to that of the ligand, it means that if the metal ion participates in the charge transport pathway, it does not significantly affect the conductance.However, it is more likely that the metal ion does not participate in charge transport, as has been found for other systems where the metal is part of an optional electron pathway. 13he similarity of the conductance values for Fe-1 and Co-1 with that of the individual ligand L 1 led us to hypothesize that the anchoring between molecule and electrode is through a single thioether contact on opposite ends of the molecule and through a single ligand (see schematic in Figure 1).Moreover, the experimental data confirm this hypothesis because of a well-defined molecular peak around 10 −3 G 0 in the histograms for all three compounds (see Figure 4a−c).These findings are supported by our transmission calculations using density functional theory (DFT) SIESTA 23 combined with the quantum transport code Gollum 24 as shown in Figures 4f  and S40.Any other configuration for the monometallic wires would result in a higher expected conductance value as shown in Figures S32−S35 in the SI.
The conductances of complexes Fe-2 and Co-2 were also measured using the STM-BJ technique.The poor solubility of these targets in nonpolar solvents meant that they could only be dissolved in higher-polar solvents.However, due to the increased lengths of the complexes and thus lowered conductance, a wax-coated tip and polar solvent could not be employed for STM-BJ measurements due to the conductance being below the leakage current threshold.To circumvent this, the complexes were preadsorbed on the surface from 1 mM solutions in ethanol for 10 min, rinsed and dried with nitrogen, and then measured in the nonpolar solvent 1,3,5-trimethylbenzene (TMB).
The measurements of Fe-2 and Co-2 proved challenging, in contrast to the clear features of the shorter wires.Not only are the conductance values (expected to be) much lower but also the junction formation probability (the percentage of binding events) is limited.In fact, only the histogram for Fe-2 shows a molecular feature at 10 −4.46 G/G 0 , see Figure 3d.We found that less than 10% of the traces for Fe-2 contained a molecular plateau, indicating that the junction formation probability is low or that the junction is not that stable.Interestingly, about a quarter of these traces contained typical "blinking" events, see Figure 5a, which are routinely observed using current−time spectroscopy in which the tip−sample distance remains constant.One possibility here is that the electrode contacts one of the thioether anchoring groups and breaks off before contacting a second thioether group next to it.However, for the Co-2 target, we observed molecular features at several different conductance values, see example traces in Figure 5c.Like Fe-2, only a small amount of the traces for Co-2 do show a plateau, but its conductance values are spread over a large range.The first two example traces (in black) show a clearly defined plateau that corresponds to the faint feature that is present in the 2D histogram in Figure 5d.However, the next three example traces (in dark purple) show less well-defined features spanning across a wide conductance range.Yet other traces only show distinct features at low and high conductance values (two example traces in magenta and lilac, respectively).All of these outcomes combined make it challenging to interpret the exact behavior of Co-2 in the junction.
Theory.We performed unsupervised clustering analysis on the data sets for Fe-2 and Co-2. 25Details of this process are provided in the SI.The algorithm separated the traces into three clusters; one cluster for Co-2 and two clusters for Fe-2 only contained clean tunneling (i.e., no evidence of junction formation).In the case of Fe-2, the remaining cluster, made of ∼26.5% of the data set, is characterized by the presence of plateaux having conductance values of ∼10 −4.5 G/G 0 and extending to lengths commensurate to the theoretical S−S distance.These can be attributed to transport through an extended configuration of Fe-2.However, for Co-2, none of the clusters show formation of junctions that extend beyond a few Ångstroms, thereby our clustering analysis does not support the idea of transport through an extended conformation of the Co-2 molecular wire.As the clustering algorithm identifies two different groups of traces, having conductances of ∼10 −1.5 and 10 −3.3 G 0 , respectively, Co-2 is either assembling in the electrode nanogap in a range of heavily tilted configurations or possibly decomposing in the junction.
The DFT calculations in Figure 4g (and Figures S36−S39 in the SI for other contact geometries) show a similar behavior for the conductance of Fe-2 and Co-2.In general, the conductance of single-molecule compounds can vary depending on the contact configuration.However, the experimental

Inorganic Chemistry
data shows that the overall conductance for Fe-2 is about 2 orders of magnitude lower than Fe-1.Comparing our experimental results with DFT calculations for single and multiple contacts (Figures S36 and S37 for Fe-2 and Figures S38 and S39 for Co-2) suggests that such a large difference takes place when we only consider one connection to the electrodes.Also, the low percentage of junction formation probability confirms that it is less likely for the molecule to bind strongly to the gold electrode through multiple connections.Therefore, we conclude that the multiple contacts are less probable.
In the case of Co-2, conductance variations with different junction geometries are also impacted by the nature of the metal centers, since the magnetic properties of cobalt cause spin polarization, which creates an extra resonance close to Fermi energy due to different resonances in spin-up and -down transmission as shown in Figure S43 in the SI.This additional resonance leads to the molecular conductance being more sensitive to the precise junction geometry.Therefore, while we could expect similar charge transport efficiencies in the two bimetallic molecular wires, the failure of Co-2 in adopting an extended configuration in the junction results in the observed very different conductance histograms.

■ CONCLUSIONS
Four new metal complexes were synthesized, with each containing three independent molecular wires bound to metal centers to form a bundle.Through the use of QCM and XPS measurements, these complexes were shown to bind to a gold surface in a fac fashion through three thiomethyl groups, leaving three unbound.However, upon measuring the single-molecule conductance of the complexes, the only detected conductance paths originated from conductance through a single ligand, indicating that both the metal ion and the other ligands had little direct impact on the conductance of the ligand.This suggests that each of the ligands maintain their independent conductance behavior in the bundle, meaning that this approach can be used to control junction geometry without impacting the behavior of the individual conductive elements.

Figure 3 .
Figure 3. XPS spectra recorded for (a) Fe-1 in the S 2p region in powder and SAM form on gold; (b) Fe-2 in the S 2p region in powder and SAM form on gold; (c) Fe-1 recorded in the Au 4f region for the uncovered Au substrate and the substrate covered by a SAM; and (d) Fe-2 recorded in the Au 4f region for the uncovered Au substrate and the substrate covered by a SAM.

Figure 4 .
Figure 4. (a, b) 1-Dimensional conductance histograms of Fe-1 (blue) and Co-1 (purple).(c) Gray shows the conductance histogram of the ligand, L 1 , used in Fe-1 and Co-1.Single-molecule conductance histograms indicate that the metal centers do participate in the dominant transport pathway.(d, e) Histograms of Fe-2 (light blue) and Co-2 (light purple).The right panels show DFT-based electrical conductance of the molecules with central atoms (f) Fe-1 and Co-1 and (g) Fe-2 and Co-2, respectively.The DFT predicted Fermi energy E F lies close to the LUMO resonance.The relaxed structures of the molecules and the junctions are shown in Figures S32, S34, S36, and S38 in the SI.In the case of cobalt, the calculation is spinpolarized and the average of the spin up and down are plotted.

Figure 5 .
Figure 5. Example conductance-distance traces along with two-dimensional histograms.(A, B) Data set for Fe-2 and (C, D) data set for Co-2.