Environmental Control of Single‐Molecule Junction Evolution and Conductance: A Case Study of Expanded Pyridinium Wiring

Abstract Environmental control of single‐molecule junction evolution and conductance was demonstrated for expanded pyridinium molecules by scanning tunneling microscopy break junction method and interpreted by quantum transport calculations including solvent molecules explicitly. Fully extended and highly conducting molecular junctions prevail in water environment as opposed to short and less conducting junctions formed in non‐solvating mesitylene. A theoretical approach correctly models single‐molecule conductance values considering the experimental junction length. Most pronounced difference in the molecular junction formation and conductance was identified for a molecule with the highest stabilization energy on the gold substrate confirming the importance of molecule–electrode interactions. Presented concept of tuning conductance through molecule–electrode interactions in the solvent‐driven junctions can be used in the development of new molecular electronic devices.


Introduction
Forl iving systems,i ti sacommon place to state that solvent (water) fully contributes to supramolecular assembling processes.A no pen issue is the extent to which this assertion remains relevant for man-made systems including cross-scale hybrid assemblies identified as molecular junction (MJ) nanodevices.W hen these functional assemblies are operating in as olvent-based environment at room temperature (contrary to ultra-high vacuum and low temperatures), such conditions are likely to impact on the active molecule functioning as the charge transporting molecular wire as well as on the energetics of contact electrodes (for example Fermi energy) and molecule-electrode interactions.This is precisely the type of assessment that we report here.
An environmental control of the charge transport in single-molecule junctions has been investigated in several recent experimental works.I na ll cases the emphasis was given on the explanation of the effect of solvent [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15] or electrolyte (conducting salt in the solvent) [16][17][18][19][20][21][22][23][24][25][26] on the conductance value of the single-molecule junction. In this contribution we will show that ac hoice of the solvent is extremely important for the junction evolution process itself, which in turn dictates the conductance value(s) obtained experimentally by the break junction methods.C hemical structure of the investigated molecule and its tendency to adsorb on the gold substrate are important factors as well.
Break junction experiments were performed in the past in the solvent environment mainly to avoid contamination and to promote the molecular junction (Figure 1a)formation. [27,28] Later as uitable solvent was used to realize electrochemical gating between several single-molecule conductance states in ON/OFF switches for molecular electronics. [20,29,30] Thee nvironmental control was also essential for the achievement of Figure 1. a) Representation of asingle-molecule junction with pyridine anchoring groups shown in grey circles. b) Chemical structures of expanded pyridinium molecules 1 to 4 having different degrees of conformational freedom around pyridinium core (q 1 , q 2 , q 4 and q 6 ). Counterions (BF 4 high rectification ratio in single-molecule diodes [4] and the notion of solvent gating was introduced. [15] Previous reports on the solvent effect consider several reasons for the change of the conductance values including as hift of the Fermi energy due to the interaction of the solvent with the electrodes [2,5,12] or ashift in the position of the transporting orbital owing to the solvent-molecule interactions. [6,15,31] Thef irst systematic study of the solvent effect was given by Fatemi et al. [2] who explained an increase of the singlemolecule conductance in thirteen different solvents by ashift of the work function of gold in contact with the solvent thus reducing the gap between the electrode Fermi energy and the energy of the charge transporting orbital. In their work the size of an investigated molecule was comparable with that of the solvents used. Authors claimed that direct intermolecular electrostatic interactions between the solvent and molecule did not play ar ole,w hereas more important was the electric dipole induced in the solvent upon its adsorption to the gold. Tr ends in the conductance changes did not correlate either with permanent dipole moments of the solvents or with their bulk dielectric constants.S olvent dependent changes of the conductance and attenuation factor b in aseries of oligothiophene-based [1] and oligoyne-based [6] molecular wires were also explained by amutual shift of the electrode Fermi energy and the transporting orbital energies,though the explanation for this shift was different from that of Fatemi et al. [2] Contrary to the previous work, Milan et al. [6] showed that for oligoyne-based molecular wires solvent-molecule interactions (solvation) alone can explain observed solvent effects. Bâldea [5] later suggested that the solvation energy,i mage charges and work function changes should be considered together to quantify the solvent effect on the molecular transport in MJ nanodevices.I nt he electrochemically gated systems the reorganization energy of the solvent is an important factor as well. [32] In such as ystem, Li et al. [8] observed temperature dependent electron transport through single redox molecules in the aqueous electrolyte suggesting astrong coupling of the redox states to water molecules.T he latest report of Tang et al. [15] stressed again the importance of the solvent-molecule interactions and solvent polarity.
Herein, we selected as eries of expanded pyridiniumbased molecules (pyridinium salts allowing the experiments in solvents of different polarity) with different degree of conformational freedom between the central pyridinium cation and the adjacent pyridine anchoring group (see Figure 1b)r anging from ap lanar system in molecule 1 to ac onformationally-locked one in molecule 3. [33] Molecules 2 and 4 have the same pattern around the pyridinium core and differ only in their molecular length. Thus molecule 4 serves mainly as ar eference compound. Synthesis and chemical characterization of these molecules are reported elsewhere. [34] It is worth noting that, owing to their appealing electrophoric and structural features (namely an easily accessible LUMO, semi-rigidity and rod-like shape), oligomers of expanded pyridiniums have already been the subject of electrochemical investigation as model molecular wires in the context of molecular electronics. [35,36] Thes ingle-molecule junction evolution and conductance Gw ere studied in selected solvents by scanning tunneling microscopy break junction (STMBJ) method that enables repeated formation and breaking of the metal-molecule contact in the molecular junction (MJ) schematically represented in Figure 1a.R etraction (current-distance) traces were converted to logarithmic conductance-distance curves and corresponding 1D and 2D conductance histograms were constructed according to procedures specified elsewhere. [37] Thereby,r elying on this STMBJ approach, we show that beyond known effects of surrounding solvent over both the active molecular component and the contact electrodes (apexes) of MJ,t his environment also sizably impacts on the conductance of MJs via their configuration, that is,b y affecting molecule-electrode interfaces.   (3, black) shifted for clarity on the Dz axis. b) 1D logarithmicc onductance (left) and 2D logarithmic conductance-distance (right) histogramsi n mesitylene(ethanol) solvent. c) 1D logarithmicc onductance (left) and 2D logarithmic conductance-distance (right) histograms in water-(ethanol) solvent. Characteristic plateau length histograms are shown in the insets.

Results and Discussion
These solvent mixtures will be hereafter referred to as mesitylene(ethanol) and water(ethanol). Based on these representative curves there is ac lear indication that the plateau length for each MJ is different.
Logarithmic conductance-distance curves show plateaus at integer multiples of quantum conductance G 0 = 77.5 mS followed by either purely tunneling current (no molecule bridging the junction) or by additional plateau(s) corresponding to the MJ conductance (see Figure 2a). Measurements in the solvents (in the absence of investigated molecules) show purely tunneling currents and were used to provide as napback distance value,w hich needs to be added to the characteristic plateau length Dz to get the experimental MJ length z exp .Further experimental details,statistical analysis of the STMBJ data for solvents and molecules 1 to 4 are given in Sections 1to4o ft he Supporting Information. Figure 2b shows statistically significant 1D logarithmic conductance (left graph) and 2D logarithmic conductancedistance (right graph) histograms for molecule 1 in mesitylene(ethanol) solvent. Figure 2c shows these 1D and 2D histograms for 1 in water(ethanol) environment. The insets in the right graphs show the characteristic plateau length Dz histograms for each solvent mixture used. Thus,the analysis of al arge ensemble of the conductance-distance curves confirms that the plateau length is shorter in mesitylene(ethanol) compared to water(ethanol) solvent, whereas the conductance of the MJ of molecule 1 in water-(ethanol) environment is higher than in mesitylene(ethanol). Theeffect of solvent on the MJ evolution was studied on the entire series of selected molecules 1 to 4 and Table 1gathers experimentally obtained single-molecule conductance values on the logarithmic scale log(G/G 0 ) exp ,experimental MJ length values z exp and junction formation probabilities in mesitylene-(ethanol) and water(ethanol) solvents obtained as an average of several data sets of retraction traces (each between 2000-4000 traces). Therepresentative 1D logarithmic conductance, 2D logarithmic conductance-distance and characteristic plateau length histograms are shown for all molecules in Sections 3a nd 4o ft he Supporting Information. In some cases,two conductance states have been found (see Figure 2b left), whereas the analysis of more prominent one was used in the following discussion. Data related to charge transport in MJ of molecule 1 in pure mesitylene have been reported elsewhere. [11] Theexperimental MJ length z exp was obtained as the most probable plateau length value Dz*i nt he characteristic plateau length Dz histogram corrected for as nap-back distance equal to 0.4 nm (see Section 2i nt he Supporting Information). Analysis of the experimental MJ length indicates that for all studied molecules,junctions break at shorter distances in the mesitylene(ethanol) as compared to the water(ethanol) environment (see Table 1).
From the junction formation probability (JP) analysis of the retraction curves (examples shown in Figure 2a)one can conclude that JP of 1 in pure mesitylene is the lowest and amounts to only 15 %a nd increases to 42 %i nm esitylene-(ethanol) and 62 %i nw ater(ethanol) solvents.T he same trend was observed for all studied molecules 1 to 4.O n average the change of the solvent from mesitylene(ethanol) to water(ethanol) almost doubles the MJ formation probability,s ee Table 1. In summary,s ingle-molecule conductance of molecules 1 to 3 is higher in water(ethanol) compared to mesitylene(ethanol), the largest difference being for molecule 1.O nt he contrary,s ingle-molecule conductance for 4 is slightly higher in mesitylene(ethanol) and follows apredicted pattern from tunneling theory,t hat is,t hat shorter MJ geometries should have higher conductances compared to longer ones.
As mentioned above,t he effect of solvent on the singlemolecule conductance was explained either by the shift of the Fermi energy due to the interaction of the solvent with the electrodes [2,5,12] or by the changing energy of the transporting orbital with respect to Fermi energy due to the solventmolecule interactions. [6,15,31] These factors can be easily incorporated within the framework of the Newns-Anderson model as was done by Bâldea. [5,38] In the present work we used the density functional theory (DFT) combined with an on-equilibrium Greensf unction (NEGF) approach to calculate single-molecule conductance values for MJ models that include explicitly solvent molecules and experimentally measured MJ lengths.T he model was developed from vacuum to that either including 6mesitylene or 42 water molecules as the solvent surrounding the expanded pyridinium molecule without further geometry restrictions.L ater,t he distance between two gold electrodes was adjusted to avalue that corresponded to the experimentally obtained MJ length z exp .A ll computational details and model development steps are given in Section 5o ft he Supporting Information. Sections 6t o1 0o ft he Supporting Information show the optimized MJ geometries,transmission functions t(e)a nd molecule-localized charge transporting orbitals and their energies for MJs of 1 to 4 in vacuum, Table 1: Experimentaland theoretical single-molecule conductancevalues expressed as log(G/G 0 ) exp and log(G/G 0 ) th ,experimental z exp and theoretical z th molecularjunction length and junction formation probability JP for molecules 1 to 4 in different solvent mixtures. [ mesitylene and water environments.F or all used model systems transmission functions were computed at zero-bias approximation and used to calculate theoretical log(G/G 0 ) th values employing Landauer formula G = G 0 t(e F ), where t(e F ) is the transmission function at the Fermi energy e F of the gold electrodes. [39,40] It is known experimentally that the position of Fermi energy e F depends on the environment (e F = À5.1 eV in vacuum). The e F of gold in contact with water is shifted by 0.6 AE 0.1 eV [5,41] and therefore the value of e F = À4.5 eV was used for water in this work.
The e F of gold in contact with mesitylene is not experimentally known and thus we decided to use the value that gives the closest agreement between theoretical and experimental Gvalues.The same approach was used previously by Milan et al. [6] Section 11 of the Supporting Information summarizes theoretically obtained single-molecule conductance values for molecules 1 to 4 (Supporting Information, Table S2) and the effect of the choice of e F value on log(G/ G 0 ) th values (Supporting Information, Tables S3 and S4).
Theoretical calculations confirmed that in all studied systems (including both environments) LUMO is the charge transporting orbital (see transmission curves in Figures 3to5) consistently with previously reported calculations for molecules terminated by pyridine anchoring groups. [42,43] Molecule localized transporting orbitals (LUMO) are shown in Sections 6to9o ft he Supporting Information.
Before assessing the solvent effect explicitly,i ti sw orth comparing intrinsic features of active molecules 1 to 4 regardless of their solvent environment. Formolecule 1 (contrary to molecules 2 and 4)LUMO does not remain confined along the main molecular axis that involves pyridine anchoring termini (see transporting orbitals in Sections 6to9ofthe Supporting Information for vacuum). In the case of molecule 1 LUMO spreads al ittle out of the longitudinal rod-like domain, over laterally fused phenyl rings.The pyridine anchor is embedded within the fused scaffold of 1 and is practically coplanar to pyridinium core ring (q 1 = 0.28 8 in Figure S5a of the Supporting Information, for the definition of q 1 see Figure 1b and Figure S5a) leading to almost fully conjugated system. [44] According to cos 2 q rule, [43,[45][46][47] the LUMO energy of 1 becomes lower compared to related species 2 and 3 containing tilted pyridine moiety (q 1 = 678 8 for 2 and 788 8 for 3, see Figures S8a and S11a of the Supporting Information). This stabilization brings LUMO energy closer to the Fermi energy e F of gold electrodes (see Table S1 of the Supporting Information) for 1 and supports the observation of the highest experimental log(G/G 0 )f or this molecule in both solvent mixtures (see Table 1). In summary,c omputational results in vacuum (see Table S2 of the Supporting Information) confirm the decrease of log(G/G 0 )going from molecule 1 to 4 in the break junction experiment.
Thes olvent effect was evaluated by the MJ model that explicitly incorporated solvent molecules into the molecular junction. Figure 3shows asummary of our theoretical results obtained for single-molecule junctions of 1 to 4 in water without any MJ length restrictions.F igure 3a shows the MJ geometry with theoretical MJ length equal to 1.3 nm for molecule 1.T heoretically obtained MJ length values z th are 1.3 nm for molecules 1 to 3 and 1.7 nm for molecule 4 in ag ood agreement with experimental MJ length values z exp (see Table 1) with only slightly higher value for the longest molecule 4.C omputed log(G/G 0 ) th values are also in very good agreement with the experiment (see Figure 3c,T able 1 and Section 11 of the Supporting Information). Figure 3b shows the corresponding transmission functions with LUMO being the charge transporting orbital. [34] We were able to reproduce our experimental results in water(ethanol) by an explicit inclusion of the water molecules and by considering ashift of the Fermi energy of gold electrodes in contact with water to av alue obtained experimentally in an independent experiment.
Thei dentical procedure was used for transmission function calculations in mesitylene solvent. In this case only 6 solvent molecules were used to keep the complexity of the system at the same level as was the case of water. Theoretical MJ length values for geometry optimized MJs of 1 to 4 in mesitylene stayed the same as those reported for water after the geometry optimization of the entire metal-moleculemetal system (compare Figure 3a and Figure 4a for molecule 1). TheMJgeometries for other three molecules are shown in Sections 7t o9of the Supporting Information. However,t he computed log(G/G 0 ) th values show much larger deviations from the experimental values as those for water environment, the largest difference being for molecule 1 (see Figure 4c).
As discussed above,i ti sk nown that single-molecule conductance values are dependent on atorsion angle between two covalently bound aromatic units. [43,[45][46][47] Applying this concept to molecule 1 we have performed transport calculations with systematically varied torsion angle q 4 between the planar pyridinium center and the adjacent pyridine anchoring group connected in a para position to the pyridinium center (for the definition of torsion angle q 4 see Figure 1b and Section 12 of the Supporting Information). In an uncon- strained system q 4 is 34.78 8 giving log(G/G 0 ) th = À2.7, which is far from the experimentally observed value À4.1 AE 0.5 (compare red bar with adotted one in Figure 4c). Theclosest agreement between log(G/G 0 ) th and log(G/G 0 ) exp was found for q 4 fixed to the unlikely value of 758 8,inwhich case log(G/ G 0 ) th was À4.0.
We have shown that the discrepancy between experimentally obtained and theoretically predicted charge transport characteristics for MJ of 1 in mesitylene solvent can be rationalized by changes of the torsion angle between planar pyridinium core and the pyridine anchor in the para position to this core.H owever,t he theoretical MJ geometry corresponds to afully extended MJ,which was not observed in the experiment performed in mesitylene.W ehave already shown (see Figure 2a nd Table 1) that the experimental MJ length values are shorter in mesitylene(ethanol) compared to those in the water(ethanol) solvent. Therefore,w eu sed these z exp values as the constrain parameter to obtain new geometry optimized MJ configurations in the mesitylene solvent for the description of the experimental data in mesitylene(ethanol) for molecules 1 to 4. Figure 5ashows the MJ configuration for molecule 1 obtained for MJ length fixed to z th = z exp obtained in mesitylene(ethanol) environment which is equal to 1.1 nm (see Table 1). TheMJconfigurations for all studied molecules at the experimentally observed MJ lengths are summarized in Section 13 of the Supporting Information. Figure 5b shows the corresponding theoretical transmission functions and Figure 5c compares the theoretical and experimental log(G/ G 0 )v alues for MJs of molecules 1 to 4.T heoretical conductances obtained after constraining the MJ length to the experimental z exp value reproduce log(G/G 0 ) exp without the need to invoke other constraints like the torsion angle discussed above for the fully extended MJ of molecule 1. Incidentally,t he angle q 4 for molecule 1 is 28.18 8 in the MJ configuration shown in Figure 5a meaning that the overall structure of molecule 1 is more planar compared to the fully extended MJ shown in Figure 4a where the torsion angle q 4 equals to 34.78 8.I ns pite of this planarity the single-molecule conductance of 1 in this MJ configuration is lower (log(G/ G 0 ) th = À4.0, see Figure 5cand Table 1) than the conductance computed for the fully extended molecular junction (log(G/ G 0 ) th = À2.7, Figure 4c). This finding seemingly contradicts the generally accepted cos 2 q rule, [45][46][47] but can be explained by as maller coupling strength between the molecule-localized transporting orbital (LUMO) and e F of the electrodes in mesitylene.The peak width of the charge transporting orbital in the transmission function is related to the coupling strength between the molecule-localized transporting orbital and e F in the Newns-Anderson model of charge transport. [48] The differences in the peak widths of the transmission functions in Figures 4b and 5b for molecule 1 (red curves) obtained by the combined DFT and NEGF approach indicate smaller coupling strength (narrower peak) in the experimental MJ geometry.
Overall, theoretical transmission curves for MJs in water (Figure 3b)c ontain much wider transmission peaks than those for MJs in mesitylene (Figure 5b)f or all studied molecules.T his means that the contact geometries at the electrode-molecule interface are indeed solvent dependent. Thed istance between nitrogen atom of the pyridine anchor and the closest gold atom of the electrode is smaller in afully extended MJs (water solvent) compared to shorter geometries (mesitylene solvent). Thus,a nother manifestation of the solvent effect stems from the solvent-induced modifica-  tion of the interaction between the molecule and gold electrodes leading to ad ifferent MJ configuration for each solvent used. One may rephrase this statement in such away that solvation effects (solvent-molecule interactions) alter the molecule-electrode interactions in the process of the MJ formation and breaking.T he importance of van der Waals forces between pyridine anchoring groups and gold substrate for the MJ evolution mechanics was already established by Aradhya et al. [49] in the absence of the solvent by simultaneous conductance and rupture force measurements.U nfortunately,w ed on ot have the computational abilities to simulate the entire MJ breaking process along the experimentally observed retraction curves using explicitly the solvent molecules.N evertheless,a fter proper consideration of the experimental MJ length we were able to explain measured MJ conductance values simply by the solvated MJ model [6] that considers solvent-molecule interactions.
Even though we are not explaining the solvent effect in terms of the individual solvent-molecule,s olvent-electrode, and molecule-electrode contributions,w ec an still assess the last contribution because we have aseries of molecules 1 to 3 with different structural arrangement around the pyridinium center and practically the same length. Calculated (in vacuum) stabilization energies of their cations on the gold-(111) substrate confirmed that adsorbed cation 1 has the highest stabilization energy followed by cations 2 and 3 (see Table S6 in Section 14 of the Supporting Information). The geometry optimized structures of these adsorbates on the gold surface (Supporting Information, Figure S19) show that nitrogen atoms of both pyridine anchoring groups are in close contact with the gold substrate in cation 1 (lying flat) and there is an increase of the distance between one of the pyridine anchors and the gold substrate going from cations 1 to 3.Thus,the role of the molecule-electrode interactions in the MJ evolution and conductance values should be most pronounced for molecule 1 as was indeed experimentally observed (see Figure 2a).
Finally,t he question as to whether aggregation [50] may occur during MJ formation is worth to be addressed. In the case of branched expanded pyridiniums (2, 3 and 4), the steric hindrance around pyridinium cores is likely to warrant the separation of wires beyond repelling of their cationic charge and solvation shell in the case of water solvent. In the most sensitive case of 4,s ingle-crystal X-ray crystallography of ac lose analogue (molecule B PP in Fortage et al. [51] )t ells us that no noticeable p-p stacking is observed when looking at crystal packing as alimiting case for solid-state organization, thereby indicating that branched species have no propensity to aggregation. Forwhat concerns the fused polycyclic species 1,c rystallography of ac lose analogue (molecule 1 H F in Fortage et al. [52] )r eveals propensity of cationic scaffolds to form stacks,which indicates that aggregation cannot be ruled out in this instance,even if rather diluted solutions (0.2 mM) were used. Noteworthy,inmore unfavorable conditions,when there is no pronounced solvation of 1 by solvent molecules that is,i nm esitylene environment, aggregate formation is likely to explain the second minor conductance peak that was not analyzed (see Figure 2b). This latter may be due to two stacked molecules facing each other and since the interaction is of van der Waals type,the conductance is lower.

Conclusion
We have observed experimentally that all studied molecules 1 to 4 form MJs with higher MJ formation probability in water(ethanol) medium than in mesitylene(ethanol) environment. Thee xperimental MJ length corresponds to af ully extended geometry in water(ethanol) and is shorter in mesitylene(ethanol) solvent. We were able to explain all experimentally observed MJ conductance values using explicitly mesitylene and water molecules and considering different MJ geometries in these two solvents.Our theoretical results (transmission functions,MJgeometries and moleculeelectrode stabilization energies) support the description of the solvent effect in which the molecule-electrodei nteractions must be taken into consideration in addition to the solvent-molecule and solvent-electrode interactions considered previously.I nv iew of ar ecent claim that high conductance transport pathway can be induced in MJs by applying potential to one of the electrodes to promote the molecular adsorption in af lat orientation, [53][54][55] our work further substantiates the importance of molecule-electrode interactions in the break junction measurements of the MJ conductance.A bove all, it evidences the critical role of surrounding medium on the MJ formation. In particular,i t shows that water favorably impacts on charge transport characteristics of cationic,r edox-active and functionally LUMO-driven electrophilic molecular wires based on expanded pyridiniums. [56]