Solvent-molecule interaction induced gating of charge transport through single-molecule junctions

To explore solvent gating of single-molecule electrical conductance due to solvent-molecule interactions, charge transport through single-molecule junctions with different anchoring groups in various solvent environments was measured by using the mechanically controllable break junction technique. We found that the conductance of single-molecule junctions can be tuned by nearly an order of magnitude by vary-ing the polarity of solvent. Furthermore, gating efﬁciency due to solvent–molecule interactions was found to be dependent on the choice of the anchor group. Theoretical calculations revealed that the polar solvent shifted the molecular-orbital energies, based on the coupling strength of the anchor groups. For weakly coupled molecular junctions, the polar solvent–molecule interaction was observed to reduce the energy gap between the molecular orbital and the Fermi level of the electrode and shifted the molec- ular orbitals. This resulted in a more signiﬁcant gating effect than that of the strongly coupled molecules. This study suggested that solvent–molecule interaction can signiﬁcantly affect the charge transport through single-molecule junctions.


Introduction
Wiring functional molecules into electrical circuits has been found to have significant potential for miniaturization of electronic devices [1][2][3][4][5][6]. Various intrinsic molecular factors have been investigated to advance the understanding on their effect on singlemolecule junction conductance, such as conformation [7][8][9], anchor groups [10,11], electrodes [12], hybridization [13], and quantum interference [14][15][16]. By altering these parameters, molecular switching can be achieved due to changing redox states [17][18][19], protonation [20,21], and connectivity [3]; however, the molecular states have been restricted to be less than three. Electrochemical and electrostatic gating [22,23] are also promising approaches for tuning the energy level alignment, although the additional gate electrode has been found to bring technical challenges for the integration. To address this issue, solvent gating has been considered a potentially simple method for continuously fine-tuning the charge transport through single-molecule junc-tions with multiple accessible states, although previous studies have reported a relatively smaller gating effect [24].
Gating of the charge transport, through molecular junctions typically occurs through electrostatic interaction [11,22]. Several mechanisms have been proposed in existing literature regarding solvent effects, such as solvent binding to the functional groups in the molecular backbone [25], solvent binding to the molecular backbone [26], solvent binding to the gold electrodes [24,27], and protonation tuning of the molecules induced by the solvent [20]. A few recent studies demonstrated a highly efficient charge transport through supramolecular interactions [28], suggesting that the electrostatic interactions between the solvent and molecule can be effectively used to tune the charge transport through single-molecule junctions.

Experimental
In this study, we demonstrated that solvent-molecule interaction resulted in a significant solvent gating effect in charge transport through single-molecule junctions, created using the mechanically controllable break junction (MCBJ) technique [29,30]. Oligophenylethynylene-sulfurmethyl (OPE-SMe) and its derivatives were employed to form single-molecule junctions [31,32]. A mixture of 20 vol% tetrahydrofuran (THF) and 80 vol%  1,3,5-trimethylbenzene (TMB) was selected as the prototype solvent for the single-molecule conductance measurement (THF for  solubility and TMB for stability and low permittivity, see SI section  1.1 for a detailed discussion on the influence of different solvents in single-molecule conductance measurement) [33]. The conductance measurements of the target molecules in different solvents were conducted using the MCBJ technique (as shown in Scheme 1). Results demonstrated that the conductance of OPE-SMe can be regulated to an order of magnitude higher by adjusting the solvent environment. This solvent effect was further investigated by varying the anchoring groups of the molecular junctions and solvent molecules.

Results and discussion
We used pure THF and TMB to prepare a 0.01 mmol/L solution of OPE-SMe, separately. In each case, conductance-displacement traces (as shown in Fig. 1a) were continuously recorded. The individual traces showed variation in the conductance as a function of distance during the process of breaking a single-molecule junction.  After a short plateau at the conductance quantum (G 0 , which equals 2e 2 /h, 77.6 Â 10 3 nS), which marked the single-atomic contact of Au-Au, the conductance trace was observed to rapidly drop to the plateau of the molecular conductance and went to the tunnelling background afterwards. The one-dimensional (1D) conductance histograms, shown in Fig. 1b, constructed from~1000 such traces, showed the conductance peaks that indicated the most probable junction conductance. The conductance of OPE-SMe, determined from the Gaussian fitting of the conductance peak, was observed to peak at 10 À2.1 and 10 À3.0 G 0 (616.4 and 77.6 nS) in THF and TMB, respectively, which showed high solventdependence. To confirm that the conductance of single-molecule junctions were measured, we constructed two-dimensional (2D) conductance-displacement histograms ( Fig. 1c and d). The dashed circles represent the overlapped area of the molecular plateaus [21]. Insets of Fig. 1c and d showed the displacement distribution analysis of the molecular plateaus [34], suggesting that the configurations of OPE-SMe junctions were stable and fully elongated in the solvents.
To further verify the effect of solvents on the charge transport of molecular junctions, conductances of OPE-SMe in various volume ratios of THF and TMB were measured. Mixed solvents with different ratios were used to gradually vary the polar environments of the system. As shown in Fig. 1e, the single-molecule conductance of OPE-SMe depended on the content of THF in the mixed solvents.
In addition, such dependence of this solvent was found in another common polar solvent (acetonitrile/ACN), as shown in Fig. 1f (detailed conductance data are shown in Figs. S2-S10 (online)), which indicated that the increase in the solvent polarity can significantly improve the charge transport of the molecular junctions.
To explore the mechanism of solvent gating on the charge transport, control experiments were carried out with oligophenylethynylene-sulfuracetyl (OPE-SAc), oligoaryleneethynylene (OAE), 1,8-octanedithiol (ODT) and 1,3-azulene (1,3Az) [2,[34][35][36][37]. Results from the conductance measurements for OPE-SAc, OAE and ODT in different solvents are shown in Fig. 2a, 2b and 2c. When compared with the results of other studies (G OPE-SAc = 10 -3.0 G 0 = 77.6 nS, G OAE = 10 -3.6 G 0 = 19.5 nS, G ODT = 10 -3.6 G 0 = 19.5 nS and G 1,3Az = 10 -3.8 G 0 = 12.3 nS), the singlemolecule conductances of OPE-SAc, OAE and ODT did not exhibit a significant solvent induced shift, while that of 1,3Az was observed to be manipulated by the solvent in the singlemolecule conductance measurement (see Figs. S11-S14 online for details). Thus, we hypothesized that the solvent effect on the molecular junctions depends on the nature of the anchor groups. As no functional groups were found on the OPE-SMe backbone for THF or TMB to form a chemical bond, and interactions between the solvents and gold atoms were too weak to affect the contact work function [24,25]. Electrostatic interactions were considered to play a dominant role in gating the conductance of molecular junctions by shifting the energy of the frontier molecular orbitals.
To further examine the response of conductance to different solvent molecules, transmission spectra T(E) were calculated by combining the DFT package SIESTA with the quantum transport code Gollum (see the Supplementary materials Section 2 for detailed method) [38,39]. The relaxed molecules, OPE, OAE and ODT, measured in the experiment were attached to the gold electrodes and were surrounded by different solvents: THF, ACN or TMB. For the solvent molecules, with an electric dipole moment (e.g., THF and ACN), we constructed conformations with dipole moments pointing towards or away from the backbone molecule, as shown in Fig. 3a and b, which show the examples of OPE-SMe with two THF molecules (additional backbone molecules with other solvent molecules are shown in the Supplementary materials). For solvent molecules without an electric dipole moment (e.g., TMB), we selected two configurations: TMB being parallel or perpendicular to the backbone, as shown in Fig. S16b (online).
To explore the role of anchor groups in the presence of polarised solvent molecule, backbone molecules were fixed during the optimization process, while the solvent molecules were allowed to relax. Results for THF are shown in Fig. 3 (see Figs. S16-S19 (online) for other backbone molecules). For OPE-SMe, when the oxygen-side of a THF solvent molecule was close to the backbone, the transmission curve was observed to shift to a higher energy (black line in Fig. 3c). In contrast, when the other end of the solvent molecule moved closer to it, the transmission curve was observed to shift to a lower energy (orange line in Fig. 3c). The two opposite shifts of the transmission function were attributed to different gat-ing effects due to the orientation of the solvent, whose dipole moment pointed from the positive to the negative oxygen-side. However, as shown in Fig. 3d-f, for the backbone molecules OAE, ODT and OPE-SAc with solvent, their transmission functions were observed to barely change over a large energy range near the DFT-predicted Fermi energy. Similar shifting trends in the transmission function of the three backbone molecules were observed in the solvent ACN (see Figs. S16-S19 online). The only difference observed between OPE-SMe and OPE-SAc was their anchor groups, which indicated that the anchor groups played a dominant role in determining the effectiveness of the solvent gating. To gain further insight into the origin of these gating effects, local density states (LDOS) around lowest unoccupied molecular orbital (LUMO), highest occupied molecular orbital (HOMO) and LUMO resonances for OPE, with anchor groups -SMe, -SAc and OAE with pyridyl anchoring groups embedded in the junction were examined, as shown in Fig. 3g-i. For anchor groups -SMe, LDOS was observed to be mainly localised on the OPE backbone, while either HOMO of anchor groups -SAc or LUMO of pyridyl anchor groups were observed to be distributed almost equally on the gold electrodes as well as on the backbone molecule (these features were also observed in the wave functions shown in Figs. S20 and S21 online). In this case, the gold and molecule were observed to be more strongly coupled and less likely to be influenced by the gating effect of the solvent. The sensitivity of the molecular orbital energy to a local perturbation, such as a solvent molecule, can be explained by the perturbation theory, which predicts that the energy shift is DE ¼ h/ n jH 0 j/ n i, where / n is the n th molecular orbital and H 0 is the perturbation. If this normalised orbital is extended over N atomic p orbitals, then its amplitude on any p orbital is ( 1= ). If the local perturbation (H 0 ) acts over M such sites, then the shift in the orbital energy DE will be proportional to M/N. Consequently, if N is large, DE would be relatively smaller, whereas, the effect would be larger for a more localised molecular orbital spread over a smaller number of sites N. This mechanism is significantly different from that mentioned in our previous work [25], where a shift in the gold contact work function was induced by solvent binding to the undercoordinated Au sites. Such a shift in the gold work function was found to be independent of the nature of the anchors.
After discussing the importance of the anchor groups on solvent gating, we will next discuss the influence of different solvents on the backbone, by taking OPE-SMe as an example. For TMB, no influence on OPE was observed in the parallel case, while a small shift was observed to appear for the standing case due to slightly positively charged H in the methyl group (see Fig. S16d online). Owing to this, we only consider the influence of THF and ACN in this section. To simulate the real experiment, 112 configurations were constructed each for OPE with one THF molecule and one ACN molecule. These curves are shown in Fig. 4a and b. Although the transmission curves could shift towards both higher and lower energies, more curves were observed to shift towards lower energies, indicating an increasing trend in the conductance when the OPE molecule was placed in a polar solvent. This was attributed to the negatively charged carbon atoms in the two phenyl rings of OPE, which attracted the positive part of the solvent molecule. These results can be clearly observed in Fig. 4c2 and c4, where more than 70% of the total configurations exhibited an increasing trend, which was observed to be consistent with the experimental results. For ACN, a larger conductance range is shown in Fig. 4c, and the conductance variations were also observed to be more sensitive to the orientation and movement of ACN, which were all attributed to the greater electric dipole momentum than that of THF. Since it is exorbitantly expensive to computationally simulate all the surrounding solvent molecules, we investigated the influence of the solvent molecules by placing just one molecule in 112 different relaxed configurations, which allowed us to collect the data [18,25,26].

Conclusions
In this study, we proposed an efficient gating approach for finetuning single-molecule conductance by controlling the solventmolecule interaction. Results showed that the conductance of the molecular wires, which weakly coupled to the electrodes, can be varied by up to 800% by adjusting the polarity of the environment. Theoretical calculations revealed that the solvents influenced the molecular orbital energies and coupling strength of the anchoring groups, and electrodes controlled the magnitude of the shift. This study demonstrated that solvent-molecule interactions can provide a highly efficient gating approach to control the charge transport in single-molecule devices. Tang carried out the break junction experiments and analyzed the data. Zheng Tang and Songjun Hou wrote the manuscript with input from all authors. Wenjing Hong, Ruihao Li and Jia Shi built the electrical measurement instrument and wrote the software to control the break junction setup. Zhibing Tan, Junyang Liu, Jueting Zheng and Yang Yang revised the manuscript. Songjun Hou, Qingqing Wu, Hatef Sadeghi, Iain Grace and Colin J. Lambert performed the theoretical modeling. All authors conceived the work and discussed the experiments.