Voltage-driven control of single-molecule keto-enol equilibrium in a two-terminal junction system

Keto-enol tautomerism, describing an equilibrium involving two tautomers with distinctive structures, provides a promising platform for modulating nanoscale charge transport. However, such equilibria are generally dominated by the keto form, while a high isomerization barrier limits the transformation to the enol form, suggesting a considerable challenge to control the tautomerism. Here, we achieve single-molecule control of a keto-enol equilibrium at room temperature by using a strategy that combines redox control and electric field modulation. Based on the control of charge injection in the single-molecule junction, we could access charged potential energy surfaces with opposite thermodynamic driving forces, i.e., exhibiting a preference for the conducting enol form, while the isomerization barrier is also significantly reduced. Thus, we could selectively obtain desired and stable tautomers, which leads to significant modulation of the single-molecule conductance. This work highlights the concept of single-molecule control of chemical reactions on more than one potential energy surface.


Supplementary Discussions
Synthetic details and characterizations 1,2-Bis(4-(methylthio)phenyl)ethan-1-one 1: To a solution of 2-(4-(methylthio)phenyl)acetic acid (10 mmol) in 50 mL dichloromethane, thionyl chloride (20 mmol) was added dropwise. A drop of dimethylformamide was also added to the solution. The reaction mixture was stirred for 1 h at room temperature. Then, the solvents were removed under vacuum. 2-(4-(Methylthio)phenyl)acetyl chloride was obtained in quantitative yield without further purification. Thioanisole (13 mmol) was dissolved in 25 mL dichloromethane at 0 ℃, and anhydrous aluminum chloride (13 mmol) was added slowly to the solution. Then, 2-(4-(methylthio)phenyl)acetyl chloride was added to the reaction mixture, and the solution was slowly heated to room temperature. After the (methylthio)phenyl)acetyl chloride was completely consumed, as detected by TLC, the reaction mixture was quenched with ice water. The organic phase was extracted by diethyl ether, which was collected and concentrated.
The crude mixture was further purified by column chromatography to give 1. Yield: 75%. White solid. 1  We first separate the total data into two groups based on the measured plateau lengths between 10 −3.2 to 10 −1.1 G0. Taking the data in 0.5 V bias as an example, as shown in Supplementary Fig. 25a and Supplementary Fig.   25c, we observe two patterns of 1D conductance histograms, which are corresponding to the high and low conductance states, respectively. The corresponding 2D conductance histograms are shown in Supplementary   Fig. 25b and Supplementary Fig. 25d. We plotted the two separated groups together with the total data in the same 1D conductance histogram. As shown in Fig. 2d, the total data is shown in the blue histograms. The blue and red curves are the profiles of the histograms of the two separated groups. We then used Gaussian fitting to fit the conductance peaks in blue and red curves, leading to the corresponding blue and red areas. in TCB under ambient conditions. c Distribution probabilities of states 'L' and 'H' are plotted against different biases. The emergence of the 'H' peak in the conductance histograms is not altered for deuterated 1-d2 compared to the regular 1 (cf. Fig. 2), indicating that proton tunneling is not a significant factor in the tautomerization mechanism. The 'H' state of 1-d2 shows a 4% lower distribution than that of 1 at 0.6 V.

Supplementary Figure 2. 1D conductance histograms of the water control experiments. a The experiment
was performed in a glovebox and a dry TCB solvent. b-e The following STM-BJ experiments were performed in the corresponding solvents under the conditions shown above the 1D conductance histograms. The control experiments (b-e) were performed under ambient conditions. TMB is the abbreviation for sym-trimethylbenzene (b). All STM-BJ experiments were performed on a 0.1 mM solution of 1. The THF we used has an initial moisture content of ~2%, which will lead to higher moisture in the solvent mixture of THF/TCB (c) than TCB itself. The water-saturated TCB was prepared by vigorously mixing with an equivalent volume of water and TCB in a glass vial. Then the water/TCB mixture was stood for 24h to get phase separation. The TCB phase, i.e., the water-saturated TCB, was carefully taken out by pipette for the following conductance experiment (d). The experiment in decane showed the same tendency (e). The emergence of the 'H' peak in the conductance histograms was not significantly affected by the presence/absence of water. Therefore, we think water may not significantly accelerate or inhibit the tautomerization process in the STM-BJ experiments.  Proving the charge injection mechanism by EC-STMBJ experiments. We have done an electrochemical (EC) STM-BJ experiment in a four-electrode system 1 to oxidize the molecule in-situ and simultaneously measure the corresponding single-molecule conductance ( Supplementary Fig. 15). The ECSTM-BJ was performed with 1.0 mM molecule 1 and 0.1 M tetrabutylammonium hexafluorophosphate as the electrolyte. The working, reference, and counter electrodes are the gold tip, Ag/AgCl, and platinum wire. In the conductance characterization, the gold tip was coated with Apiezon wax 2 to reduce background capacitive current and electrochemical currents. As shown in Supplementary Fig. 15a, the cyclic voltammetry characterization was started at −0.2 V with an oxidative scan by the ECSTM-BJ setup. In the oxidative scan, the cyclic voltammetry of molecule 1 shows two consecutive oxidation peaks at around 1.0 V and 1.3 V EC potentials (relative to Ag/AgCl), which correspond to the one-electron and two-electron oxidization and suggest that oxidation states are accessible. We observed three reduction peaks at around 0.3, 0.7, and 1.2 V in the reductive scan. The reductive peaks at 0.7 and 1.2 V should be the redox pairs corresponding to oxidation peaks at 1.0 and 1.3 V. The new reductive peak that appeared in the reductive scan at 0.3 V suggests that there was a significant structural reorganization in the oxidative scan.
We use an EC gate and measure the single-molecule conductance simultaneously. All the conductance characterizations are performed with 0.1 V bias applied between the tip and substrate. As shown in Supplementary Fig. 15b, the 0 V EC gate measurement reveals a mono-conductance state around 10 −4.5 G0, which is the low-conductance state (keto form) of molecule 1. At the 0.3 V EC gate ( Supplementary Fig. 15c), there was only the signal of the low-conductance state. When applying a 1.0 V EC gate, we observe two distinctive conductance peaks, centering around 10 −2.0 G0 and 10 −2.8 G0 (Supplementary Fig. 15d). Its 2D conductance histogram indicates that the conductance plateaus of 10 −2.8 G0 are about two times longer than the plateaus of 10 −2.0 G0 (shown in Supplementary Fig. 15f), while the two types of plateaus occur consecutively.
The conductance plateaus of 10 −2.8 G0 are very similar to the high-conductance state (enol form) of the twoelectrode measurement at 0.6 V (Fig. 2d). We also observe that upon two-electron oxidation, a low-conductance state is reached again (shown in Supplementary Fig. 15e). This state most likely corresponds to a deprotonated keto form. To understand the structure in oxidation, we also measured the UV/Vis spectra of the oxidized species. We expect that the enol form will have a more extended pi-system than that of the corresponding keto form.
Therefore, the enol form will have a smaller bandgap than that of the keto form. As expected, the methylated enol form 1-OMe (with the extended π-system) exhibits a redshift absorption (50 nm) with respect to molecule 1 ( Supplementary Fig. 16, grey curve). More importantly, after one one-electron oxidation in molecule 1, we observed a similar redshift absorption ( Supplementary Fig. 16, red curve), which is consistent with the absorption of 1-OMe. The UV/Vis spectra further support that the enol form was generated from the keto form upon one- Excluding the possibility of connectivity change. To exclude the possibility that conductance switching comes from connectivity change, we perform a series of control experiments. As shown in Supplementary Fig. 17a, we synthesized reference molecules R1 and R2 with one -SMe anchor removed compared to molecule 1.
Subsequently, we characterize the single-molecule conductance of these molecules by STM-BJ. As shown in Supplementary Fig. 17b and 17c, the 1D conductance histograms do not show any clear conductance peak with the bias ranging from 0.1 V to 0.6 V, which is in sharp contrast to the result of molecule 1 shown in Fig. 2d. The above results suggest that the oxygen in molecule 1 cannot contact gold electrodes to form stable and observable junctions even with a higher bias applied. If the high-conductance state were originating from a shorter junction geometry, we would expect to observe two consecutive conductance plateaus, suggesting a positive correlation between the high-and low-conductance states. To examine whether such a correlation exists, we performed a correlation analysis. As shown in Supplementary Fig. 28b, the correlation matrix for the dataset of 0.6 V bias shows a clear negative correlation (red areas) between the high-and low-conductance values, which means that the two conductance plateaus occurred in a mutually exclusive pattern. As a reference, such a negative correlation does not emerge from the dataset at 0.1 V bias ( Supplementary Fig. 28a). Tip-speed related state switching probability. As indicated in the manuscript, tautomerization involves a nonnegligible reaction barrier that needs to be crossed, and consequently, the switching takes some time to transpire.

Supplementary
As a result, the tip speed can be expected to affect the ratio of the high-and low-conductance states. We performed some additional experiments to confirm this reasoning. As shown in Supplementary Fig. 18a, when we change the tip speeds in the STM-BJ experiments with 0.5 V bias applied, the ratio of high-conductance states indeed decreases significantly as the tip speed is increased from 5 to 20 nm s −1 . In the 2D conductance histograms, the high-conductance plateaus in Supplementary Fig. 18b are much more significant than in Supplementary Fig. 18c. We are considering that there is about a 0.5 nm stretching distance before the contact of a molecular junction breaks down. Thus, we could estimate the junction lifetime by dividing the stretching distance by the tip speeds. As shown in Supplementary Fig. 18d, we find that the high-conductance state ratio is increased with the longer junction lifetime. The above results indicate that the amount of transport charge is also vital for the induced tautomerization.

Supplementary Note 2. Valence Bond interpretation of the impact of oxidation on the PES.
In this Supplementary note, we will take a closer look at the root cause for the dramatic difference in tautomerization behavior of the uncharged and charged species of the molecular bridge. The electron transfertriggered switch in thermodynamic preference can be straightforwardly understood by simply considering some common bond dissociation enthalpy (BDE) tables found in the literature 3 . To simplify the analysis and facilitate the discussion, we start by considering the tautomeric system in the absence of thiol-linkers. The potential energy profiles associated with tautomerization (with opposing electric field directions, cf. the main text) for this system are presented in Supplementary Fig. 19.  Fig. 20a). As such, the thermodynamic driving force ∆ associated with this reaction can be estimated at +8 kcal mol −1 (i.e., 98 + 105 -110 -85). Note that this crude estimate agrees perfectly with the calculated value presented in Supplementary Fig. 19.
In the case of charged species, it is straightforward to see that -in the idealized situation that the electron is removed exclusively from the keto-/enol-moiety of the molecule under consideration -there is only a single active bond in the tautomerization process: the HRR'C-H bond is broken in the reactant and replaced by the RO-H one in the product ( Supplementary Fig. 20b). Using the same bond dissociation enthalpy values as for the uncharged species, 1 one ends up with an estimated thermodynamic driving force ∆ of −5 kcal mol −1 . As such, we readily recover the experimentally observed thermodynamic switching, i.e., ∆ goes from +8 → −5 kcal mol −1 upon removing an electron from the system. However, it should be clear that the obtained driving force does not correspond numerically to the calculated value presented in Supplementary Fig. 19a, i.e., The reason for this discrepancy is that the unpaired (radical) electron in the cation can readily delocalize, and will do so to a much greater extent in the enol-form than in the keto-form, due to the conjugation of the lone pair on the oxygen moiety with the adjacent -type C-C •+ system (cf. Supplementary Fig. 21a). The fingerprints of this resonance delocalization can be seen among others in the bond lengths: both the C-C and C-O bond lengths in the enol are intermediate between a single and double bond ( Supplementary Fig. 21b), similar to what one observes for a prototypical delocalized allyl radical. As some of us discussed in detail in a recent contribution, a loss or gain of delocalization is inherently connected to a relative thermodynamic penalty paid throughout the reaction 4 . In the tautomerization reaction under consideration, it is the enol-form that is resonance stabilized compared to the keto-form. Based on the preceding argument, the magnitude of the resonance penalty associated 22 with the tautomerization reaction can thus be crudely estimated at 17 kcal mol −1 , i.e., the magnitude of the discrepancy between the naively estimated and the actual, calculated ∆ .
Supplementary Figure 21. Orbital analysis. a The delocalized -system in the charged enol; b the bond lengths for R, R' = Ph.
Next to the switching in thermodynamic preference, the dramatic change in the kinetics of the tautomerization upon oxidation can also be readily understood from a qualitative valence bond (VB) analysis 5 . Central to VB reactivity theory is the construction of so-called valence bond state correlation diagrams (VBSCD), which depict the evolution of individual diabatic states, corresponding to the electronic configuration of the reactant and the product, along the reaction coordinate. In practice, these diabatic states interact and mix as the reaction proceeds, and collectively they give rise to the adiabatic curve, i.e., the full ground-state PES associated with the chemical reaction under consideration ( Supplementary Fig. 22).
Within this VB framework, the approximate barrier height associated with a generic chemical reaction can be estimated with the help of the following expression, (1) where corresponds to a fraction ( ) of the average of the promotion gap on the reactant and product side, Supplementary Equation (2) Let us now estimate the barrier height associated with the tautomerization reaction for both the uncharged and charged species. For the neutral species, a schematic VBSCD with the main representative VB structures for the reactant and product diabatic curves in respectively the reactant and product geometry are shown in Supplementary Fig. 23a.
The (vertical) promotion gap between R and R* can be estimated upon inspection of these two structures. In essence, R can be turned into R* by unpairing the RR'C=O and HRR'C-H bonds and at the same time re-pairing the resulting unpaired electrons residing on the adjacent carbon atoms into an RR'C=CHR" bond (the O and H centers are too far away in the reactant geometry for them to exhibit an actual bonding interaction). The energy required to "(un)pair" the electrons involved in a bond can be expressed as the corresponding singlet-triplet excitation energy. EST values for each of the bonds broken/formed during the promotion from R to R* and from P to P* respectively. The subscripts "R" and "P" denote whether the reactant or product geometry was used in the single-point calculations; the subscripts "broken" and "formed" denote whether the bond is an existing bond that is being broken during the promotion event, or a new bond that is being formed during the promotion.
bond ΔEST (kcal mol −1 ) due to symmetry reasons, mixing of the individual VB structures is limited in such a situation; taking cyclobutadiene (the most "notorious" square system) for example, B has previously been determined to amount to approximately 21 kcal mol −1 . 6 We can now follow the same approach to estimate the barrier for the charged species (cf. the VBSCD in Supplementary Fig. 19b). With the help of Supplementary   Fig. 24). The higher extent of localization in the positive field means that a bigger gain of delocalization energy is associated with the tautomerization reaction towards the enol under these circumstances.
Hence, the thermodynamic driving force increases, and according to the Bell-Evans-Polanyi principle (as well as Supplementary Equation (1)), this will also reduce the reaction barrier. 2 For the actual molecular bridge with S-anchors included, the same analysis as above is valid. There is however one major complication: upon charging, the Sulfur-moieties carry a significant spin density, i.e., the positive charge is delocalized a lot more than without the anchors. As a result of this charge delocalization, the dramatic effect on thermodynamics and kinetics, observed in Supplementary Fig. 19, is somewhat tempered.
Whereas the uncharged PES is not altered meaningfully upon inclusion of the anchors, i.e., the barrier still amounts to approximately 60 kcal mol −1 and the keto-form is more stable than the enol-form by approximately 10 kcal mol −1 , the charged PES exhibits a significantly smaller thermodynamic driving force towards the enolform (7 kcal mol −1 vs. 22 kcal mol −1 ) and a significantly higher reaction barrier (35 kcal mol −1 vs. 23 kcal mol −1 ).
Supplementary Equation (1) indicates that the bulk of this barrier height increase (7.5 out of 12 kcal mol −1 ) is caused by the reduced thermodynamic driving force, i.e., the change in the extent of delocalization; the rest most likely stems from a change in the B-factor.

Supplementary Note 3. Assessment of the feasibility of the charge transfer event.
To gauge whether a thermal transition from the uncharged to charged Marcus parabola is feasible (cf. Fig. 5b in the main text), and can consequently explain the tautomerization is giving rise to the observed switching behavior in the conductance, we followed the analysis by Nitzan and co-workers. 7 Let us first consider the thermodynamic driving force, i.e., the energy difference between the minimum of the two parabolas in Fig. 5b. This energy difference, Δ , can be expressed as follows: Note that these values agree within reasonable bounds with the calculated thermodynamic driving force towards the enol-form on the charged PES (see main text).
Let us now take a look at the potential existence of a thermal barrier separating the uncharged and charged keto-species. According to Marcus theory, the barrier height for a charge transfer event can be estimated to amount to, Supplementary Equation (4) where corresponds to the reorganization energy. In accordance with the methodology used by Nitzan and co-workers, 7 the reorganization energy was estimated from gas-phase calculations on the isolated molecular bridge: the energy difference between and was determined in the optimal geometry of the uncharged molecular bridge, after which the resulting value was corrected to reflect the actual alignment between the two Marcus parabolas as inferred from the experimental data, i.e., Δ = 9.2 -12.7 kcal mol −1 (cf. Supplementary Equation (3)). Using this approach, approximate values of 11.3 -14.8 kcal mol −1 could be determined. Plugging these values into Supplementary Equation (4), one ends up with a barrier height in the range of 9.3 -12.8 kcal mol −1 .
As such, the inferred thermodynamic driving force together with the crudely estimated reaction barrier suggests that a transition from the uncharged parabola to the charged parabola can occur at room temperature.
Hence, this thermal charge transfer mechanism is a plausible trigger of the tautomerization reaction; the bias window does not need to reach the HOMO peak in the transmission spectrum for the electron transfer to occur.

Supplementary Note 4. Estimation of the alignment of the Fermi level with the HOMO transport channel.
In order to fit the experimental I-V curve, we solve the following equation 8 (6) and T(E) is approximated as a single resonance channel, i.e., the HOMO eigenchannel, centered at energy  through a Lorentzian function, Supplementary Equation (7) By varying the distance between the resonance and the Fermi level ( − ), as well as the resonance width , one can tune the shape of the calculated I-V curve. An optimal agreement between the calculated and measured I-V curve is reached at − ≈ 0.7 and = 0.012 (see Fig. 3c in the main text).  Fig. 4a. The contacting pads were connected to gold wires (0.1 mm in diameter) through the silver conductive paint (SCP03B, Electrolube).