Charge Transport Across Dynamic Covalent Chemical Bridges

Relationships between chemical structure and conductivity in ordered polymers (OPs) are difficult to probe using bulk samples. We propose that conductance measurements of appropriate molecular-scale models can reveal trends in electronic coupling(s) between repeat units that may help inform OP design. Here, we apply the scanning tunneling microscope-based break-junction (STM-BJ) method to study transport through single-molecules comprising OP-relevant imine, imidazole, diazaborole, and boronate ester dynamic covalent chemical bridges. Notably, solution-stable boron-based compounds dissociate in situ unless measured under a rigorously inert glovebox atmosphere. We find that junction conductance negatively correlates with the electronegativity difference between bridge atoms, and corroborative first-principles calculations further reveal a different nodal structure in the transmission eigenchannels of boronate ester junctions. This work reaffirms expectations that highly polarized bridge motifs represent poor choices for the construction of OPs with high through-bond conductivity and underscores the utility of glovebox STM-BJ instrumentation for studies of air-sensitive materials.


Synthesis and Characterization
All manipulations were carried out in oven-dried glassware under a nitrogen atmosphere using standard Schlenk line techniques. No special precautions were taken to exclude air or moisture during workup unless otherwise stated. Tetrahydrofuran (THF) and dichloromethane (CH2Cl2) were sparged with nitrogen and dried using a two-column solvent purification system packed with alumina (Pure Process Technologies, Nashua, NH, USA). N,N-Dimethylformamide (DMF) was purified by vacuum distillation, dried over 3Å molecular sieves, 1 and stored under nitrogen. Ltd, Manchester, UK) to leave only the apex area exposed. 5

Measurements under Inert Atmosphere
Air-free STM-BJ measurements were performed using an ambient temperature and pressure STM-BJ setup identical to the one described above but now housed inside of a customized OMNI-Lab 4-port glovebox (Vacuum Atmospheres Company, Hawthorne, CA, USA; Figure S1).
Experiments were typically started at single digit or sub-ppm H2O and O2 concentrations, as measured by internal glove atmosphere sensors, though the limits of operation have not been systematically probed. O2 concentrations typically increase to 5-10 ppm during an experiment, as estimated by sensor readings upon re-initiation of air circulation (see below). To the best of our knowledge, previous single-molecule junction experiments conducted under inert atmosphere have utilized environmental chamber, glove bag, or glovebox-based environmental controls that are typically less effective at maintaining low H2O and O2 conditions, or these conditions have not been explicitly reported. [6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21] To isolate our setup from acoustic, electrical, and vibrational noise, the STM head was enclosed in a grounded aluminum box lined with acoustic foam. This environmental isolation box was supported by a conductive aluminum plate which in turn was suspended from the interior ceiling of the glovebox using elastic cords. Neodymium magnets positioned beneath the suspended plate provided additional eddy current dampening. BNC and USB feedthroughs (GS GLOVEBOX Systemtechnik GmbH, Malsch, Germany) were used to connect the STM head to controller and measurement hardware outside the box. Gold substrates were UV-ozone cleaned immediately prior to pumping/sparging them into the glovebox and used for measurements that same day.
Solvents for air-free STM-BJ measurements were sparged with nitrogen and dried over 3Å molecular sieves 1 prior to use. To minimize noise during data acquisition, circulation of the glovebox atmosphere was suspended, circulation valves were closed to isolate the catalyst, the vacuum pump was switched off, and the glovebox gloves were tied down to provide additional atmospheric stabilization. Figure S1. Schematic of glovebox STM-BJ setup: (a) The glovebox is capable of maintaining an nitrogen atmosphere with <1 ppm H2O, <1 ppm O2 (light blue); (b) Large and (c) small vacuum/nitrogen antechamber for bringing items in and out of the glovebox; (d) Glove ports enable manipulations within the glovebox; (e) STM head environmental isolation box; (f) Hooks connect a conductive aluminum base plate to the glovebox interior ceiling via elastic cords (green); (g) A sunken chamber floor containing an adjustable lab jack (dark blue) allows the STM-BJ head to be raised (providing stabilization to make experimental adjustments) and lowered (for vibrational isolation). The jack is topped with neodymium magnets (purple) to provide eddy current damping at small plate-magnet distances; (h) Interior electrical cables (blue dashed lines) connect to (i) external cables and hardware via KF-40 feedthroughs (red).

Transmission Calculations using DFT+Σ
The junction structures were constructed by placing seven Au (111) layers (with 4×4 Au atoms on each layer) at each side of the molecule, with the sulfur atom in the molecule binding to an Au trimer. 4,22 During the geometry relaxation, the outer three Au layers on each side were considered as a rigid body, with their relative positions kept as those in the bulk and the force acting on these three Au layers taken as the average force on the atoms in the fourth Au layer. All degrees of freedom of the extended molecule (the inner three Au layers on each side + an Au trimer on each side + the molecule) are fully relaxed, until all forces are below 0.04 eV/Å. The geometry relaxation used the SIESTA package, 23 the Perdew-Burke-Ernzerhof (PBE) functional, 24 a 4×4×1 k-mesh. The Au pseudopotential and basis functions were adapted from prior work 25 and were S6 chosen to reproduce the work function of Au (111) surface. Single-zeta basis functions were used for Au, and double-zeta basis functions were used for all other elements.
After junction geometry relaxation, the transport calculations were performed within the framework of non-equilibrium Green's function as implemented in TranSIESTA, 26 using the same functional, pseudopotential, basis set, and k-point sampling. After the non-equilibrium density matrix is converged, the coherent transmission functions as a function of energy were computed using the Landauer formula in a post-processing manner, 27 with a 16×16 k-mesh. To correct the quantitative errors of PBE functional in estimating the transport properties, the DFT+ Σ approach 25,28 was applied following the procedure outlined by Liu et al. 22 In this approach, the self-energy correction to the molecular orbitals (defined as the eigenvalues in the subblock of the molecular Hamiltonian within the junction) includes two contributions: a gas-phase contribution and a surface polarization, which we discuss separately below.
The gas-phase contribution, Σ gp , is calculated for the isolated molecule. For occupied EA is the electron affinity of the molecule, defined as E(N)-E(N+1). Both IP and EA are calculated using total energy differences. As a result, all quantities on the right-hand side of the equation above are -dependent, and the OT-RSH HOMO or LUMO energies are those evaluated at the optimal that minimizes ( ). In Table S1, we report the values, HOMO and LUMO energies calculated from both OT-RSH and PBE for the gas phase molecule, as well as the Σ gp . All gasphase calculations were performed using the Gaussian 16 package 30 with a cc-pVTZ basis.
The surface polarization, Σ img , is modelled using a classical image-charge model, 28 where the image-plane position is set to be at 1.47 Å above the top Au (111) layer (excluding the trimer binding motif). The charge density of the molecule is approximated by a collection of point S7 charges located at the nuclei with the charge being the Mulliken charge of each atom. In Table   S1, we also report the Σ img for HOMO and LUMO, as well as the total self-energy correction, Σ = Σ gp + Σ img . We assume all occupied levels have the same self-energy correction and all unoccupied levels have the same self-energy correction.

Evaluation of HOMO-LUMO Gaps of 1D Oligomers
For a complete picture of how DCC bridge groups modulate the delocalization of frontier orbitals, we probe the p-conjugation in 1D molecules that are analogous to the model compounds studied in the main text. The structures of the 1D molecules are shown in Figure S15. We optimize the molecular structures using the B3LYP functional with the 6-31++G** basis, using the Gaussian 16 package. To enable a direct comparison between the 1D molecules studied here and the model compounds studied in the main text, we have constrained all molecular structures to be planar during the geometry optimizations (the free optimization would have resulted non-planar structures for some of the molecules in the trans-2CN and cis-2CN series). The trends in HOMO-LUMO gaps are summarized in Figure S15, with orbital plots of the longest oligomer in each series presented in Figure S16 and the raw data of HOMO and LUMO values for all molecules compiled in Table S2.

4-(Methylthio)benzene-1,2-diamine (MeS-2N)
This method was adapted from an analogous procedure. 31 DMF (60 mL) was added to a flask containing 5-chloro-2-nitroaniline (4.491 g, 26.02 mmol) and sodium thiomethoxide (2.98 g, 42.5 mmol). The mixture was heated to 65°C with vigorous stirring for 20 h. After cooling to room temperature, the dark-red mixture was transferred to a separatory funnel and extracted with ethyl acetate (2 × 200 mL). The combined organic extracts were washed with water (3 × 50 mL) and brine (50 mL) before drying them over anhydrous MgSO4. After filtration, solvent was removed by rotary evaporation and the residue purified via flash chromatography (SiO2; ethyl acetatehexanes, 0:1→1:1 v/v) to provide 1 as a dark red oil that darkened and solidified overnight (1.813 g, 38%; stored under a nitrogen atmosphere at 5°C). Spectroscopic data was consistent with previous reports. 31  1 (1.011 g, 5.488 mmol) and anhydrous SnCl2 (6.314 g, 33.30 mmol) were added to a threeneck round-bottom flask equipped with a reflux condenser. Ethanol (40 mL) was added, and the reaction mixture was heated at reflux for 20 h. After cooling to room temperature, the pH of solution was adjusted to 10-13 by careful addition of aqueous 3 M NaOH. The mixture was then diluted with ethyl acetate (400 mL) and filtered through Celite. The filtrate was washed with brine (50 mL) then dried over MgSO4. After filtration, solvent was removed by rotary evaporation to provide MeS-2N as a brown solid (0.805 g, 95%). Spectroscopic data was consistent with previous reports. 31

4-(Methylthio)benzene-1,2-diol (MeS-2O)
This method was adapted from analogous procedures. 32  then slowly cooled to room temperature. The yellow crystalline solid that precipitated was collected by filtration and washed with hexanes to provide 2CN (1.158 g, 85%). Spectroscopic data was consistent with previous reports. 35

CN-Me
This method was adapted from an analogous procedure. 37

BN
This method was adapted from an analogous procedure. 38

2CN-L
This method was adapted from the one used to prepare 2CN, described above. Terephthalaldehyde

CN-L
This method was adapted from the one used to prepare CN, described above. A mixture of terephthalaldehyde (0.138 g, 1.03 mmol), MeS-2N (0.315 g, 2.04 mmol), and sodium metabisulfite (0.581 g, 3.06 mmol) in DMF (4 mL) was heated to reflux overnight with vigorous stirring. After cooling to room temperature, deionized water (20 mL) was added. The resulting dark-green precipitate was collected by filtration and purified by flash chromatography (SiO2; ethyl acetatehexanes, 1:1 →0:1 v/v) to provide CN-L as a brown solid (0.086 g, 21%). NMR spectra show partially overlapping resonances for three possible tautomers (only one tautomer is shown in the reaction scheme for simplicity). 1

BN-L
This method was adapted from the one used to prepare BN, described above.

BO-L
This method was adapted from the one used to prepare BO, described above. Ethyl acetate (~14

Stability of Boronate Ester and Diazaborole Junctions
The stability of BO and BN junctions when measured in the absence of air (Figure 2c In (a, c), the solution was measured in the glovebox under inert conditions, showing a persistent well-defined peak at ~10 -4 G0 . In (b, d), a sample of the same solution was removed from the glovebox and immediately subjected to measurements in air. A similar peak feature is observed in histograms built from the first 200-500 traces, which then rapidly decreases in intensity over ~30 min. Both sets of measurements were performed on the same day. Figure S5. Overlaid 1D conductance histograms for (a) BN and (b) 2CN, measured in air and in a nitrogen-filled glovebox (Vbias = 100 mV, 5,000-10,000 traces). No significant differences are observed between the (primary) conductance peaks when measuring these molecules in air or under an inert atmosphere.

Apparent Noise in Glovebox STM-BJ Conductance Histograms
Beyond the interpretation of molecular conductance trends, we also recognize that the histograms obtained from glovebox-based STM-BJ experiments presented in Figure 2c exhibit an apparent higher noise than for analogous measurements in air, as indicated by the increased counts between 10 -2 -10 0 G0 and decreased resolution of atomic point contact features above ~1 G0 (see also Figure   S6). While our glovebox STM-BJ setup operates in a more challenging chemical, vibrational, and acoustic environment with the potential to increase measurement noise, we find that these apparent noise features are typically absent in clean gold measurements prior to the addition of solvent (see Figure S7a-c, yellow). Glovebox STM-BJ studies in different anhydrous, deoxygenated solvents reveal that the apparent noise features are present when using TCB and mesitylene, but absent with tetradecane (TD; SI, Figure S7a-c, brown). These features are not observed for measurements in air using solvent samples taken from the same batches as used in the glovebox measurements, ruling out solvent contamination during drying or deoxygenation. We hypothesize that such "noise" features are attributable to solubilized adventitious impurities (e.g., present on the gold surface or in the atmosphere) or increased interactions between undercoordinated gold atoms and TCB/mesitylene in the absence of air. Notably, the presence of oxygen has been found to increase the length of single atomic gold chains in low temperature break-junction measurements, 43 suggesting that oxygen (and/or water) may serve to stabilize undercoordinated gold atoms during STM-BJ measurements in air. Figure S6. Expanded view of overlaid 1D conductance histograms for TCB solutions of 2CN measured in air and in a nitrogen-filled glovebox (Vbias = 100 mV, 5,000-10,000 traces). We observe increased counts between 10 -2 -10 0 G0 (arrows) and decreased resolution of atomic point contact features above ~1 G0 (starred peaks). Figure S7. Top: Overlaid 1D conductance histograms obtained for clean gold substrates measured in a nitrogen-filled glovebox before (yellow; ≥1,000 traces) and after (brown; ≥5,000 traces) addition of (a) TCB, (b) mesitylene, or (c) TD (Vbias = 100 mV). Bottom: Overlaid 1D conductance histograms obtained for clean gold substrates measured in air before (yellow) and after (brown) addition of the same samples of (d) TCB, (e) mesitylene, or (f) TD (identical Vbias and number of traces as for (a)-(c)). The lack of significant apparent noise features for clean gold measurements measured in the glovebox prior to solvent addition (a-c, dark yellow) shows that the operation of this instrument is not significantly affected by chemical contamination, or vibrational or acoustic noise. For glovebox measurements using TCB or mesitylene we observe increased background counts and decreased resolution of atomic point contact features, relative to measurements of the same gold substrates prior to solvent addition, or to measurements using TD.

Predictions from Chemical Models
Application of the "extended curly arrow rules" recently presented by O'Driscoll and Bryce 44 to CN-m (where the C=N bond is positioned meta to -SMe, Figure 3a) suggest that it should exhibit a "shifted destructive" quantum interference (SDQI; Figure S8a) rather than DQI. Here the electron-donating para-connected -NH-substituent (absent in 2CN-m) is considered to increase the conductance of this tautomer/isomer relative to that of 2CN-m by moving the transmission antiresonance away from the Fermi level (EF). While the predicted result from this model agrees with our findings for CN-m, hybridization between molecular states and gold d-states complicates interpretation of the change in transmission in terms of a simple shift in antiresonance position ( Figure S12). We also note, for completeness, that further arrow pushing allows us to draw a zwitterionic resonance form for CN-m that places the C=N bond para to the -SMe group (SI, Figure S8b), resulting in an electronic structure that resembles that of the most important resonance form of CN-p. Figure S8. (a) Application of "extended curly arrow rules" 44 to the "CN-m" tautomer having the C=N bond meta-connected to one thioether (-SMe) group (denoted here as A = acceptor). The lone pair of the electron-donating -NH-group positioned para to the same -SMe/A substituent can be delocalized through curly arrow electron pushing to both -SMe/A. This indicates that CN-m should exhibit "shifted destructive" quantum interference (SDQI, transmission antiresonance far from EF) rather than DQI (transmission antiresonance near to EF). (b) A zwitterionic resonance form for CN-m that places the C=N bond para to both -SMe groups.

Measurements in a Polar Solvent (Propylene Carbonate)
Control studies in propylene carbonate (PC) show that changes in solvent or bias polarity (at low biases) do not significantly impact the most probable conductance of the molecular junctions studied here (Figure S9)      . While only minor differences are observed between HOMO-LUMO gaps for trans-and cis-2CN oligomers, the gap decrease for meta-BO oligomers is different to para-BO and the smallest of all oligomers studied. We attribute the distinct properties of meta-BO to the meta-connectivity in these oligomers which limits orbital delocalization. Figure S16. Isosurface plots of frontier orbitals for the longest oligomers shown in Figure S15 as illustrative examples. The HOMO and LUMO for 2CN oligomers extend over 3-5 bridge groups without interruption. Nodes on atoms limit orbital delocalization across ~3 or fewer groups for BO oligomers, most significantly for meta-BO.              Figure S38. Figure S26. 13 Figure S38. Figure S28. 13 Figure S38. 1 H NMR (400 MHz) spectra of (a) BN and (b) BO measured at different time internals over 1 week in toluene-d8 (a chemically comparable aromatic solvent), to probe their hydrolytic stability during STM-BJ experiments in TCB solutions. NMR tubes were uncapped in-between measurements to maximize air exposure. The spectrum for MeS-BO (a common hydrolysis product) is provided in each panel for comparison. Resonances attributable to MeS-BO are absent, or do not appreciably increase in intensity (where present as minor impurities), in the BN and BO solution samples. This shows these compounds do not significantly hydrolyze in solution after 1 week even in the presence of adventitious water (starred peaks at ~0.43 ppm). ~ toluene-d8 ~ ~ ~ toluene-d8 ~ ~ Figure S39. 1 H NMR (600 MHz) spectra for a sample of BO in ~0.6 mL toluene-d8 measured before (top) and after (bottom) stirring open to air for ~2 h with 0.05 g Au powder (<10 μm diameter) in the presence of adventitious water (starred peaks at ~0.43 ppm). 42,51 Prior to spectroscopic characterization, the stirred solution sample was passed through a sand-cotton wool filter to remove Au powder. No appreciable spectral changes are observed, indicating that significant molecular dissociation of the solute to soluble, stable products does not occur upon contact with a gold surface area of approximately the same order of magnitude as present in STM-BJ studies, on relevant measurement timescales ( Figure S4). ~ toluene-d8 ~