Low Vapor Pressure Solvents for Single-Molecule Junction Measurements

Abstract

Nonpolar solvents commonly used in scanning tunneling microscope-based break junction measurements exhibit hazards and relatively low boiling points (bp) that limit the scope of solution experiments at elevated temperatures. Here we show that low toxicity, ultrahigh bp solvents such as bis(2-ethylhexyl) adipate (bp = 417 °C) and squalane (457 °C) can be used to probe molecular junctions at ≥100 °C. With these, we extend solvent- and temperature-dependent conductance trends for junction components such as 4,4′-bipyridine and thiomethyl-terminated oligophenylenes and reveal the gold snapback distance is larger at 100 °C due to increased surface atom mobility. We further show the rate of surface transmetalation and homocoupling reactions using phenylboronic acids increases at 100 °C, while junctions comprising anticipated boroxine condensation products form only at room temperature in an anhydrous glovebox atmosphere. Overall, this work demonstrates the utility of low vapor pressure solvents for the comprehensive characterization of junction properties and chemical reactivity at the single-molecule limit.

Nonpolar solvents such as 1,2,4-trichlorobenzene (TCB) or 1-phenyloctane (PO) are widely used in scanning tunneling microscope (STM) experiments for surface self-assembly^1,2 or single-molecule conductance³⁻⁵ studies. The moderately high boiling points (bp) of these and other commonly used nonconducting compounds (Table 1) mitigate their evaporation at room temperature (RT) and minimize Faradaic and capacitive currents that complicate measurements in higher polarity media.⁶ While the solvent environment is key to achieving sufficient analyte dissolution, it can also control interfacial charge transport and chemical reactivity. In STM-based break junction (STM-BJ) experiments, solvents comprising heteroatoms with a relatively high binding affinity for gold have been found to influence the energetic level alignment between the dominant conducting orbital(s) of the molecule and the Fermi level (E_F) of the electrodes, rendering junction tunneling conductance solvent-dependent.⁷⁻¹⁰ One such solvent, 1-bromonaphthalene (BN), can even influence the most probable binding geometry of gold–thiol junctions by competing for undercoordinated surface sites.¹¹ Should charge hopping processes dominate transport, the solvent environment around the junction may also impact conductance by modulating outer sphere reorganization energies associated with sequential electron transfer events.¹²⁻¹⁴ The solvent dielectric can also tune the strength of interfacial electric fields shown to catalyze in situ chemical transformations.^15,16

Table 1. Selected Properties of Nonpolar STM-BJ Solvents Explored in This Work^a.

^aChemical structures are shown in Figure 1 and Figure S2. Heading abbreviations: mp = melting point, bp = boiling point, LD50 = median lethal dose. Table rows are shaded to indicate key solvent characteristics: red = commonly used, hazardous; orange = infrequently used, hazardous, higher bp; green = underutilized, nontoxic, ultrahigh bp.

^bFrom Material Safety Data Sheets, the amount of an orally administered material which causes the death of 50% of a population of rats.

^cNotable hazards include H = hazardous (skin irritant/sensitizer, toxic to organs, or may be fatal if swallowed) or A = toxic to aquatic life.

^dHazards not fully characterized.

Critically, most nonpolar solvents do not exhibit sufficiently low vapor pressures to inhibit significant evaporation at elevated temperatures or over extended periods at RT.⁹ To minimize the impact of varying solution concentration, previous studies have measured samples at high temperatures prior to low temperature⁵ or have periodically added pure solvent to analyte solutions.^16,17 We propose that new, robust strategies to minimize solvent loss could enable several useful studies and comparison experiments, including (1) precise control of reagent concentrations to probe the kinetics of in situ chemical reactions, (2) use of high temperatures to drive in situ chemical reactions that are unfavorable at RT, (3) rigorous assessment of thermally triggered molecular switches,¹⁸ or (4) the evaluation of single-molecule charge transport mechanisms (e.g., tunneling vs hopping).^7,19 It has been suggested that more studies of molecular junctions at >85 °C, the likely operating temperature of hybrid molecule–semiconductor integrated circuits, are needed to evaluate their potential technological applications.²⁰ We also note that the volatilization of hazardous solvents such as TCB through extended use or deliberate heating is a potential health hazard. This may prove particularly problematic in typical low airflow laboratory environments housing STM instrumentation, where acoustic isolation/electromagnetic shielding chambers concentrate solvent vapors. As such, there is a clear need to identify ultrahigh bp, low toxicity solvents for STM studies, particularly for experiments at elevated temperatures.

In this work, we introduce bis(2-ethylhexyl) adipate (DEHA) and squalane (SQ) as safe, low vapor pressure solvents that facilitate STM experiments at temperatures of ≥100 °C over extended periods (≥1 d). We first demonstrate that both N lone pair and pyridyl π-coordinated 4,4′-bipyridine (bipy) junctions successfully form from DEHA and SQ solutions at RT.²¹ We subsequently show that junctions formed from thioether-terminated oligophenyls in DEHA at 100 °C exhibit not only an increased conductance compared to RT, corroborating a recent report,⁵ but also a decreased junction step length. Snapback measurements indicate that the change in step length may in part be attributed to the increased mobility of gold at elevated temperatures that results in faster electrode restructuring after breaking atomic point contacts. Finally, studies of phenylboronic acids in DEHA show that the rate of previously reported in situ C(sp²)–C(sp²) bond forming reactions²² is significantly increased at 100 °C relative to RT. Despite driving off water in studies at ≥100 °C, we find no evidence for boronic acid condensation reactions in air, in stark contrast to glovebox STM-BJ measurements performed at RT in an anhydrous nitrogen atmosphere which are dominated by the boroxine cyclotrimerization product. Taken together, we anticipate the methods and findings introduced in this study will prove critical for evaluating the potential impact of time and elevated temperature on the conductance, stability, and function of molecular-scale junctions formed from solution.

In Table 1 we first consolidate key metrics for established and new nonpolar STM-BJ solvents. Entries 1–5 comprise common STM-BJ solvents that may reasonably be considered to exhibit high bp: TCB,³⁻⁵ 1,3,5-trimethylbenzene (mes),²³ tetradecane (TD),²⁴ BN,^7,12 and PO.⁵ However, we find that the small quantities of TCB typically used in STM experiments (25 mg) completely evaporate at 100 °C after 1 h (Supporting Information, Solvent Evaporation Studies). This rapid loss of solvent imposes experimental limitations as indicated in a recent study of the temperature-dependent conductance of oligophenylene wires, in which the maximum temperature reported for measurements in TCB was 55 °C (in PO, 73 °C).⁵ Notably, all these commonly used solvents exhibit specific hazards as reported in Material Safety Data Sheets obtained from commercial suppliers (red rows, Table 1).

We sought to identify other compounds with melting points of <25 °C and bp of >271 °C. While dibutyl maleate (mal), hexadecane (HD), and 1-octadecene (OD) qualify, they still pose significant hazards (orange rows, Table 1). In contrast, DEHA and SQ fall into a distinct category, having both ultrahigh bp (≥417 °C) and low toxicity (green rows, Table 1). While the environmental health impact of DEHA is still under active evaluation, it is a commonly used plasticizer in food packaging with >14 500 tons produced annually in the U.S. and EU combined.²⁵ A branched alkane with formula C₃₀H₆₂, SQ is a naturally occurring component of human sebum and widely used in cosmetics.²⁶ Remarkably, we find that small quantities (29–84 mg) of DEHA or SQ can be heated to 100 °C for 1 d with <4% mass loss. This indicates these solvents could be used to perform STM studies at ≥100 °C, a milestone temperature for exploring chemical dynamics and reactivity under ambient pressure. Such experiments could facilitate useful comparisons with conventional solution-based variable-temperature nuclear magnetic resonance spectroscopic studies and have the potential to drive in situ condensation reactions by generating an anhydrous environment that shifts the position of associated chemical equilibria.

We evaluated the utility of each new solvent to form and study single-molecule junctions using previously described custom-built STM-BJ instrumentation (see Supporting Information for additional details).^3,27,28 In the STM-BJ method, a gold STM tip is repeatedly pushed in and out of a gold substrate while applying a bias voltage (V) between the tip and substrate electrodes and using the measured current (I) to calculate conductance (G = I/V). Initial large area gold junctions thin with increasing tip–substrate displacement to produce single-atom point contacts, as indicated by step features in conductance-displacement traces with an average conductance of ∼1 G₀ (=2e²/h). In the presence of molecules that can bridge the electrodes, we form single-molecule junctions after breaking the point contact that result in additional step features at conductance values <1 G₀ (Figure 1a). We compile thousands of consecutively measured traces into conductance histograms without data selection, whereby the most probable conductance values and other characteristic features can be analyzed. For variable-temperature STM-BJ studies, we apply a custom-built resistive heating stage that controls the temperature of the substrate only (Figure S1). At temperatures above RT, STM-BJ experiments are not easily conducted without solvent as the tip undergoes rapid thermal expansion and contraction when moved in and out of substrate contact. When both tip and substrate are in contact with solvent, they reach thermal equilibrium in <15 min.

Figure 1.

(a) Left: Schematic illustration of a 4,4′-bipyridine (bipy) single-molecule junction, formed through solution self-assembly using different solvents. Right: Molecular structures of commonly utilized 1,2,4-trichlorobenzene (TCB) and previously unreported bis(2-ethylhexyl) adipate (DEHA) and squalane (SQ) solvents (boiling points noted in parentheses). (b) Overlaid 1D conductance histograms for bipy measured in TCB, DEHA, and SQ, each showing the double peak feature characteristic of these junctions (>8000 traces, V_bias = 100 mV).²¹ (c) Corresponding 2D conductance histogram for bipy measured in DEHA. Conductance data for bipy measured in the other solvents listed in Table 1 are presented in Figures S3 and S4. (d) Plot of high (square) and low (triangle) conductance peak features in different solvents (see Figure S5 and Table S2 for details of peak fits). Junctions formed in solvents containing heteroatoms (solid) typically exhibit a lower conductance than solvents comprising only C and H atoms (hollow).

We initially targeted studies of bipy at RT which exhibits two conductance peaks associated with binding through different junction geometries.²¹ The high conductance feature is attributed to a tilted pyridyl-gold binding configuration that enhances electrode-pyridyl π-system coupling. The low conductance feature is associated with a perpendicular pyridyl-gold binding configuration that reduces coupling to the π-backbone orbitals. In Figure 1b, we plot representative overlaid histograms for bipy measurements in TCB, DEHA and SQ. Data for the other solvents are provided in Figure S3. Each histogram clearly shows the characteristic features of bipy junctions and the absence of excessive background noise attributable to Faradaic and capacitive charging currents, confirming all solvents in Table 1 are suitable for STM-BJ experiments.

These studies also reveal that the high and low conductance features of bipy junctions (G_bipy) vary by a factor of ∼2 when measured in different solvents (Figure 1d and Table S2) and that G_bipy is typically lower in solvents comprising heteroatoms (Cl, Br, O). Following previous reports we suggest that these solvents, which are expected to bind more strongly to undercoordinated gold atoms, can effectively displace bipy from surface sites in the vicinity of the bipy junction.¹⁰ This should reduce the surface dipole and decrease the local Fermi level (E_F). For bipy junctions, where conductance is dominated by transport through the lowest unoccupied molecular orbital (LUMO),^21,29 the increased energy difference between E_F and E_LUMO will result in a decrease in G_bipy. This trend, as anticipated,⁹ is opposite to that observed for 1,4-benzenediamine junctions where conductance is dominated by transport through the highest occupied molecular orbital.¹⁰

To benchmark DEHA and SQ as STM solvents for studies at elevated temperature, we first performed solution measurements of a thioether-terminated terphenyl wire (P3SMe, Figure 2a, inset). The tunneling conductance of junctions formed from such oligophenyl wires in TCB and PO solutions has been reported to increase with increasing temperature up to 73 °C, attributed to the thermally enhanced planarization of the oligophenyl backbone.⁵ In Figure 2a, we present overlaid conductance histograms for P3SMe obtained in DEHA at RT (blue; 22–28 °C) and 100 °C (red) that corroborate these findings. However, we also observe a noticeable decrease in the intensity of both molecular junction and gold point contact peak features. Similar results are obtained for P3SMe measurements in SQ (Figure S6), and for measurements of analogous (PnSMe) wires in DEHA (n = 2,4; Figure S8a,c). Analysis of the corresponding 2D conductance–displacement histograms and junction step-lengths for P3SMe shows that this lower intensity junction peak results from a step feature that is 0.25 nm shorter (Figure 2b–d).

Figure 2.

(a) Overlaid 1D conductance histograms (5000 traces, V_bias = 100 mV) obtained for a 1 mM solution of P3SMe (structure inset) in DEHA at room temperature (RT, blue) or 100 °C (red). At elevated temperatures, the molecular conductance peak shifts to higher conductance and decreases in intensity. (b, c) 2D conductance histograms corresponding to the histograms shown in (a). These show that the reduced intensity for molecular junction and gold point contact (∼1 G₀) result from shorter step features. (d) Overlaid step length histograms with Gaussian fits (solid lines), obtained from the data sets used in (a)–(c). The average difference between junction step lengths is 0.25 nm. (e) In snapback measurements, a conductance-displacement trace (blue, left axis) is obtained with a piezo ramp (black, right axis) applied to move the tip out and back in to contact with the substrate in DEHA (no analyte molecules). L1 is the tip–substrate displacement between breaking the junction (G < 0.9 G₀) and reversing the piezo direction, L2 is the displacement required to reform the junction (G > 0.9 G₀). (f) Overlaid snapback distance histograms with solid line fits for L2–L1 (the gold snapback distance) at RT (blue, 0.76 nm) and 100 °C (red, 1.16 nm). The difference in snapback between RT and 100 °C is 0.40 nm, which is comparable to the average difference in junction step lengths (d).

Given that the reduced intensity 1 G₀ peak at elevated temperature has been attributed to the increased mobility of gold surface atoms,³⁰⁻³² we hypothesized that shorter junction steps may be rationalized by an increased gold snapback distance under these conditions. This distance has been previously measured at room^21,23,33,34 and cryogenic^35,36 temperature(s) and results from the rapid restructuring of the gold interface to a lower energy configuration after rupturing the point contact. As junction step lengths are typically found to correlate with the length of the molecular component after subtraction of the snapback distance,³³ we expect the step length to decrease as the snapback distance increases. In Figure 2f, we present overlaid snapback distance histograms from snapback measurements in DEHA at RT and 100 °C (see Figure 2e and the Supporting Information for experimental details).³⁴ Peak fits confirm the average snapback distance at 100 °C in DEHA is 0.4 nm larger than at RT, providing at least a partial explanation for the shorter junction step lengths measured under these conditions (discussed further in Figure S7). A similar observation has been reported for Ag junctions at lower temperatures.³⁷ We propose this phenomenon was not recognized in previous studies of PnSMe at 53–73 °C ⁵ as at these temperatures in different solvents the differences in snapback distance may be smaller or the junction formation process may be distinct (Figure S7).³³

We further demonstrate the utility of these low vapor pressure solvents by applying them to probe the interfacial reactivity of phenylboronic acids exposed to gold surfaces at different temperatures and to explore whether their condensation to cyclotrimerized boroxine products could be driven through the elimination of H₂O at elevated temperatures (Figure 4a).³⁸ Building on reaction chemistry demonstrated using nanoparticle or supported gold catalysts,^39,40 Li et al. have shown that thioether-terminated phenylboronic acids and esters are capable of undergoing transmetalation processes at the solution–electrode interface to form aryl-Au linked junctions during STM-BJ measurements.²² These adsorbed species were found to subsequently undergo surface-mediated C(sp²)–C(sp²) homocoupling processes, as has also been demonstrated in temperature-programmed desorption⁴¹ and STM studies^42,43 involving gold substrates and halogenated precursors.

Figure 4.

(a) Scheme showing the formation of BOr from P1B through a condensation reaction. This equilibrium can be driven to the right by the removal of water. (b) Overlaid 1D conductance histograms for BOr and P1B measured in anhydrous DEHA under an inert nitrogen atmosphere (10 000 traces, V_bias = 750 mV). (c) 2D histogram corresponding to the BOr measurement in (b). (d) Overlaid 1D conductance histograms for BOr and P1B measured in anhydrous TD under an inert nitrogen atmosphere (10 000 traces, V_bias = 750 mV). Together, these measurements reveal a new conductance feature around 1 × 10^–5G₀ that is attributable to the formation of BOr junctions in the absence of water at RT. Additional 2D histograms are provided in Figure S13.

In Figure 3a we plot overlaid conductance histograms for 4-methylthiophenylboronic acid (P1B) measured in DEHA at RT and 100 °C. At RT our histograms are comparable to those previously reported from measurements of P1B in TCB,^22,28 showing a clear feature at 1 × 10^–2G₀ (“(i)”) but only a broad peak at 6 × 10^–4G₀ (“(ii)”). Upon heating to 100 °C, peak (ii) sharpens and increases in intensity, and we observe a new feature at 4 × 10^–6G₀ (“(iii)”) that corresponds to a step that forms after breaking the step associated with peak (ii) (Figure 3c). Following earlier reports,²² we attribute peak (i) to the Au–C bound transmetalation product and peak (ii) to junctions comprising the homocoupled biphenyl species (Figure 3b). These assignments are further corroborated by the observation of comparable peak features in studies of 4-thioanisole-AuPPh₃ (comprising a preinstalled aryl-Au bond).⁴⁴ As indicated in Figure 3b, we attribute peak (iii) to π-stacked dimer junctions of P2SMe, noting that this feature is also observed in measurements of pure P2SMe in DEHA (Figure S8). While measurements of P1B were performed with a tip–substrate bias of 750 mV to resolve peak (iii), the intensity of peaks (i) and (ii) similarly increase upon heating to 100 °C at a bias of 100 mV (Figure S11). This demonstrates that temperature is a particularly effective handle for driving such chemical reactions.

Figure 3.

(a) Overlaid 1D conductance histograms (10 000 traces, V_bias = 750 mV) obtained for measurements of P1B solutions in DEHA measured at RT (blue) and 100 °C (red). At elevated temperatures, a low intensity feature observed in RT measurements at 6 × 10^–4G₀ (ii) becomes more pronounced, and a new feature is observed at 4 × 10^–6G₀ (iii). (b) Proposed molecular junctions formed from P1B, following (i) transmetalation of the phenyl group to gold, (ii) homocoupling of the transmetalated product to form P2SMe; (iii) π-stacking between two P2SMe molecules. (c) 2D histogram for the measurement of P1B at 100 °C, corresponding to the 1D histogram in (a). (d) Overlaid 1D histograms (5000 traces, V_bias = 750 mV) for P2B measured in DEHA at RT (blue) and 100 °C (red). (e) Proposed molecular junctions formed from P2B. (f) A plot of conductance versus n for measurements of PnSMe prepared ex situ (hollow/dotted) and in situ (solid) at RT (blue) and 100 °C (red). Ex situ conductance values are an average of ≥2 measurements, and in situ values are taken from a single measurement. Conductance and β values obtained from fits to G = G_ce^(−βn), as well as additional conductance histograms, can be found in Table S3 and Figures S8–S10.

Comparable changes in peak features upon heating are observed for 4-(4-methylthiophenyl)phenylboronic acid (P2B), which we attribute to formation of the corresponding transmetalation product and P4SMe (Figure 3d,e). We also observe a feature that we attribute to P3SMe upon heating a mixture of P1B and P2B in DEHA (Figure S8e). In Figure 3f, we overlay plots of the most probable conductance vs n for measurements of PnSMe prepared ex situ and in situ at RT and 100 °C. This analysis reveals that PnSMe junctions prepared in situ exhibit a greatly reduced temperature-dependence compared to junctions formed from the ex situ synthesized species. Figure 3f also shows that the conductance of PnSMe junctions prepared in situ is typically lower than that obtained for measurements of the wires synthesized ex situ using solutions comprising pure analyte material. While further studies are needed, we suggest that these differences may be attributed to the presence of chemisorbed Au-aryl species proximal to the measured junction which could hinder the enhanced planarization of phenyl rings at elevated temperatures and shift the local E_F further from the dominant conducting orbital(s) of PnSMe.¹⁰

Remarkably, conductance histograms of P1B obtained from measurements at 100 °C, or even 110 °C (Figure S12), in air contain no peak features that may be assigned to the formation of the boroxine product BOr (Figure 4a), likely due to rapid transmetalation. To verify this observation, we synthesized BOrex situ and subjected it to glovebox STM-BJ measurements in anhydrous DEHA at RT under an inert nitrogen atmosphere. The corresponding conductance histograms comprise a prominent new feature around 1 × 10^–5G₀ that we attribute to the formation of BOr junctions (Figure 4b,c). While a peak at ∼1 × 10^–2G₀ is still apparent, the absence of a feature attributable to P2SMe junctions at ∼1 × 10^–3G₀ indicates the peak around 1 × 10^–5G₀ may not be attributed to the π-stacked P2SMe dimer geometry (Figure 3a,b). Measurements of P1B under the same conditions yield comparable conductance histograms albeit with a more prominent feature at ∼1 × 10^–3G₀. This indicates that, in contrast to measurements at ≥100 °C, BOr readily forms in situ from solutions of P1B prepared in an anhydrous environment at RT which serves to shift the equilibrium in favor of the condensed species (Figure 4a). We note that the conductance of BOr junctions is comparable to analogous components comprising 1,3-connected triazine or benzene central rings (2 × 10^–5G₀ and 6 × 10^–6G₀, respectively, in tetrahydrofuran/mes).⁴⁵ The lack of clear correlations between conductance and the difference in electronegativity of backbone heteroatoms in this family of molecules, observed for a variety of other compounds,^28,46,47 likely reflects a complex interplay of orbital (de)localization and destructive quantum interference in these cross-conjugated systems.^45,48,49

These experiments appear to demonstrate that P1B can undergo the same interfacial chemistry under an inert nitrogen atmosphere as those observed during measurements in air, whereas control studies^22,50 and proposed mechanisms^22,51 suggest O₂ is key to activation of the boronic acid group. Critically however, our glovebox STM-BJ studies are typically performed in the presence of O₂ at ppm concentrations,²⁸ meaning we cannot rule out the presence of adventitious oxygen adsorbed at the electrode–solution interface. DEHA also comprises ester groups which could feasibly interact with gold in a similar manner to that proposed for O₂. To help evaluate this hypothesis, we present, in Figure 4d, conductance histograms obtained from glovebox STM-BJ measurements of BOr and P1B in anhydrous TD. These present analogous conductance features to those obtained from studies in DEHA, while peaks attributable to boronic acid activation do appear reduced in intensity. However, it remains difficult to deconvolute other influences of the solvent environment on these interfacial reactions, such as the reduced solubility of analytes in TD compared to DEHA or the reduced polarity of TD which may destabilize charged intermediates.

Here we have shown that low toxicity, ultrahigh bp solvents such as DEHA and SQ can be used to routinely perform STM-BJ measurements in solution at 100 °C. In proof-of-concept studies involving prototypical analytes such as bipy and PnSMe, we apply these solvents to extend previously reported solvent- and temperature-dependent conductance trends. Critically, we also show that measurements at 100 °C can result in an increased snapback distance and significantly reduced molecular junction step lengths. In demonstrating that temperature can be applied in STM-BJ studies to drive and study chemical reactions, we show that the oligophenyl products formed in situ from the coupling of aryl boronic acids exhibit distinct conductance properties to the isolated species synthesized ex situ. We anticipate that the new capabilities provided by these or other ultrahigh bp solvents will strongly motivate additional investigations that target single-molecule devices or interfacial chemistry controlled through temperature modulation or extended molecule–electrode contact. As demonstrated here, such studies can provide new insights into junction characteristics and surface-based reaction mechanisms critical to the effective exploitation of these nanoscale functional assemblies and interfacial processes.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.nanolett.4c02786.

• Additional experimental details, solvent evaporation studies, synthetic methods, and conductance data (PDF)

Author Contributions

^† L.K. and T.M.C.-B. contributed equally to this work.

The authors declare no competing financial interest.