Control over Molecular Orbital Gating and Marcus Inverted Charge Transport in Molecular Junctions with Conjugated Molecular Wires

Recently it is discovered that molecular junctions can be pushed into the Marcus Inverted region of charge transport, but it is unclear which factors are important. This paper shows that the mechanism of charge transport across molecular wires can be switched between the normal and Marcus Inverted regions by fine‐tuning the molecule–electrode coupling strength and the tunneling distance across oligophenylene ethynylene (OPE) wire terminated with ferrocene (Fc) abbreviated as S‐OPEnFc (n = 1–3). Coherent tunneling dominates the mechanism of charge transport in junctions with short molecules (n = 1), but for n = 2 or 3 redox reactions become important. By weakening the molecule—electrode interaction by interrupted conjugation, S‐CH2‐OPEnFc, intramolecular orbital gating can occur pushing the junctions completely into the Marcus Inverted region. These results indicated that weak molecule—electrode coupling is important to push junctions into the Marcus Inverted Region.


Introduction
Achieving control over charge transport (CT) between molecules and electrode surfaces is fundamental for the creation of new device materials in areas ranging from organic electronics [1] and catalysis [2,3] to energy management [4] and biology. [5] For this reason, it is important to study the mechanisms of CT across molecule− electrode interfaces in detail. Molecular junctions make it possible to study the mechanisms of CT at the smallest possible length-scale under very well-controlled conditions giving unprecedented insights into, for example, thermoelectric properties, [6] light-matter interactions, [7] and switching mechanisms. [8] Recently, the Marcus Inverted region (MIR) for CT has been found in redoxactive molecular devices where charges traverse electrode-molecule-electrode junctions via incoherent tunneling (also called hopping) without a noticeable thermally activated component (activation energy, E a ) to cross the hopping barrier. [9,10] Unlike redox processes in solution, [11] in junctions the energy from the leads compensates for E a . Recently, it has been shown that this elimination of E a in the MIR can be used to improve the performance of molecular devices. [12] Incoherent tunneling has been frequently observed in molecular junctions but MIR has been only recently reported for a very few systems [9,10,12,13] and it remains unclear which factors push molecular junctions into MIR because the mechanism of CT depends on several intertwined factors. Here we show that MIR can be turned on and off by varying the coupling strength of the molecule with the electrode (via the insertion of a single methylene -CH 2 -unit) which works only for junctions with long, redox-active molecular wires. On the other hand, coherent tunneling dominates in short wires. Unraveling the interplay between molecular length and molecule-electrode coupling strength and understanding how this interplay affects the mechanism of CT (incoherent vs coherent tunneling) is important for future designs of molecular junctions.
In essence, the Landauer and Marcus theories describe CT at the two opposite ends of the spectrum. The Landauer-Buttiker theory describes coherent tunneling and is widely used to model current flow across molecular junctions with high coupling factor between the molecules and electrodes. [14,15] The Marcus theory describes incoherent tunneling (also called hopping) and is widely used to explain incoherent redox processes in solution. [18,19] In molecular tunnel junctions, in principle both processes can take place but it is still poorly understood when a junction transits from coherent to incoherent tunneling because of the difficulty to disentangle the factors. As we show here, molecular length, moleculeelectrode coupling strength, and applied voltage all affect the CT mechanism.
The Landauer-Buttiker model works very well for short molecules with strong molecule-electrode coupling, but in this work, we show that for molecules with increasing length and decreasing molecule-electrode coupling (low γ), redoxreactions become important, and the Landauer model alone is not suitable.
For junctions where redox processes in the molecules cannot be ignored, such as in this work, alternative theories involving the Marcus theory have been proposed that describe incoherent redox processes in solid-state junctions that lack an electrolyte. [20][21][22] Migliore and Nitzan developed a model combining Marcus and Landauer theories that successfully explains CT in junctions where the relative distance between the Marcus parabolas is affected by electrical gating between the involved charge states. They achieved this by decoupling the slow redox transitions between the neutral (M) and charged (M + ) states of a molecule inside junctions (modeled by the Marcus theory) from the fast sequential tunneling of electrons/holes between the electrodes and the respective molecular charge states (modeled by the Landauer theory). [23] According to this model, Equations 1 and 2 describe the transport rates to (K M M → +) and from ( K M M → + ) the charged states.
Here Γ(E) is the tunneling rate between the molecular state and the electrodes, ΔE = ΔE 0 + μ is the electrochemical potential of the electrodes where ΔE 0 is the energy difference between the two molecular states ( ), x M and x M + are the molecular configurations at the equilibrium of M and M + where x is the reaction coordinate, μ is the electrochemical potential of the leads, and λ is the reorganization energy. The model allows the inclusion of an internal electrical gating potential which affects the activation-free energies of the fluctuation probabilities of x M and x M + and thereby shifts the relative energy position of the Marcus parabolas of the initial and the final states with respect to one another, leading to an effective change in the activation energy E a . This internal gating results from the interaction between the different charge states of the molecule as the charge inside the molecule increases with increasing bias. This effective gating potential is directly proportional to the charge in the molecule (which depends on bias) and capacitive coupling between the two charge states. It can be written as Here Q represents the bias-dependent electron charge on the molecule and capacitance C C represents the effective capacitive coupling between the donor and the acceptor which is independent of V but determined by the molecular structure. The activation energy E a , is given by Equation (3) and is thus dependent on V g .
Molecular junctions that operate in MIR are very rare and it is unclear under which conditions junctions operate in MIR or Marcus normal regions. The vast majority of known redoxactive molecular junctions operate in the Marcus normal region, but, as exemplified in our previous work, in junctions with a donor-bridge-acceptor (D-b-A) compound, specifically S-CH 2 -DPA-(CH 2 ) n -Fc (n = 1−3, DPA = diphenylacetylene and Fc = ferrocene), charges can hop from the donor (Fc) to the acceptor (DPA). [9] During this hopping process, the Fc units are charged, Fc + , resulting in a lowering of the energy of the acceptor level with respect to that of the donor due to intramolecular orbital gating. Figure 1c illustrates this effect in terms of Marcus parabolas for charge transfer from the donor to the acceptor and shows that for sufficiently large intramolecular orbital gating, the parabola of the initial (neutral) state Fc-DPA intersects one of the final (zwitterionic) state Fc + -DPA − at its minimum producing barrierless charge transport. Beyond this point, the system enters the MIR and nonzero E a is restored which in principle should slow down the electron transfer rates, but in molecular junctions this energy penalty is compensated by the energy from the electrodes. Recently, we showed that the Marcus parabolas can be shifted with respect to each other because the different charge states of a redox molecule interact differently with the applied electric field. [10] This mechanism can also explain the results of Kang et al. [24] who observed MIR for a molecule with only an electron acceptor group. However, for all these examples the mechanism of charge transport strongly depends on the intertwined factors of the moleculeelectrode coupling strength (γ) and the molecular length (d) and so it remains unclear why they operate in the MIR instead of the Marcus normal region.
In this work, we address the question of how MIR can be accessed in molecules where the donor and acceptor groups are fully conjugated and, in principle, intramolecular orbital gating strength should be strong enough to observe MIR. A fully conjugated molecular wire may couple too strongly to the electrodes resulting in coherent, rather than incoherent, tunneling diminishing MIR. To investigate the role of both γ and d in the mechanism of CT of redox-molecules operating in the Marcus Inverted region, we incorporated a conjugated oligophenylene ethynylene (OPE) of variable length functionalized with Fc, S-(Ph-CC) n -Fc (S-OPE n Fc, n = 1-3), or alternatively a molecular wire with an interrupted conjugation at the thiol anchoring group S-CH 2 -(Ph-CC) n -Fc (S-CH 2 -OPE n Fc, n = 1-3), in EGaIn junctions ( Figure 1). Thus, we control γ via the -CH 2moiety and d via n. The major conclusion of this work is that with a single -CH 2 -unit we can reduce γ enough to push the molecular junctions into the MIR, but only for junctions with n > 1. Junctions lacking the -CH 2 -unit, and those with short molecules (d < 1.5 nm for n = 1), remain in the strong coupling regime where coherent tunneling dominates the mechanism of charge transport.

Characterization of the Monolayers
To characterize both the electronic and supramolecular structure of the SAMs, we used a range of characterization techniques (see Section S1-S7, Supporting Information) and the SAM properties are summarized in Table 1. From these results, we make the following observations. i) The surface coverages determined by cyclic voltammetry (CV) are within error the same for all OPE derivatives and are significantly higher than the previously reported value for S-OPE 1 Fc monolayers on Au electrode (1.2 × 10 −10 mol cm −2 ) [26] but lower than for SAMs made of only OPE units (9.7 × 10 −10 mol cm −2 ). [27] ii) The monolayer thickness d SAM determined from angle-resolved X-ray photoemission spectroscopy (XPS) scales with the number of OPE units. The S-CH 2 -OPE n Fc monolayers are 2-3 Å thicker than S-OPE n Fc monolayers with the same n due to the extra -CH 2unit. iii) The energy of the HOMO (E HOMO ), determined by EGaIn stands for eutectic gallium-indium alloy and GaO x represents the 0.7-1.0 nm-thick gallium oxide layer. [25] The straight arrow indicates coherent tunneling and the curved arrows indicate incoherent tunneling (or hopping). The double arrows indicate the strong (solid black) and weak (dashed black) coupling between OPE and the Au electrode, and the intramolecular orbital gating (red) between the donor D (Fc) and acceptor A (OPE). c) The Marcus parabolas for the electron transfer between Fc and OPE. The red arrow indicates the intramolecular orbital gating effect, which moves the parabola of the D + −A − state down into the activationless (dashed red line) and then the Marcus Inverted regime (solid red line). ultraviolet photoemission spectroscopy (UPS), is within error the same for all derivatives, but the shift of E HOMO with respect to E F , δE ME,H = E HOMO -E F , increases with n. iv) The optical HOMO-LUMO gap, E g,o , determined with UV-Vis spectroscopy, decreases with n. Consequently, the energy of the LUMO, E LUMO-UV , decreases with increasing n. v) The decrease of δE ME,L = E LUMO -E F , was verified with near-edge X-ray absorption fine structure (NEXAFS) spectroscopy. vi) The tilt angle of the SAM with respect to the surface normal is 4-9° smaller for the S-CH 2 -OPE n Fc SAMs than for the S-OPE n Fc SAMs. From these observations, we conclude that all SAMs form dense monolayers with the molecules in an upright position with a very similar overall supramolecular structure of the SAMs except that S-CH 2 -OPE n Fc SAMs are less tilted (and therefore more ordered) than S-OPE n Fc SAMs. The overall electronic structure of all the SAMs is remarkably similar but δE ME,L decreases by ≈200 meV, and δE ME,H increases by roughly 200 meV with increasing n ( Figure S33, Supporting Information).
To elucidate in more detail how both types of SAMs couple to the Au surface, we performed density functional theory (DFT) calculations. Figure 2e shows that for S-OPE 1 Fc, the LUMO is completely delocalized with contributions from the Fc, OPE, sulfur, and metal. By contrast, in S-CH 2 -OPE 1 Fc, the CH 2 linker has almost no contribution to the LUMO and breaks the conjugation between OPE and the metal surface. Based on this result, we conclude that S-OPE n Fc and S-CH 2 -OPE n Fc SAMs can be considered as strongly and weakly coupled SAMs. In addition, the calculations show the HOMO of the systems are localized on the iron atom in Fc (Figure 2f) with minor contributions from the sulfur atom. For more details, see Figure S29-S32, Supporting Information. Thus, these SAMs are highly asymmetrical and are potential rectifiers analogous to ferrocenylalkanethiolates. [28] Since these monolayers are new, we discuss their electrochemical behavior in more detail. Peak deconvolution (Figure 2c,d) shows that the cyclic voltammograms are dominated by the main peak representing the Fc groups in a wellpacked SAM. Only a small post-peak that likely originates from physisorbed molecules is present, similar to what we have reported for ferrocenylalkanethiolate SAMs. [29] The contribution of this post-peak diminishes with increasing n indicating that long molecules form better-packed SAMs than short molecules, likely due to the larger supramolecular OPE-π-π network for long molecules. The peak oxidation potential is very similar for all SAMs (Table 1). Interestingly, the full width at half maximum (FWHM) values for the main anodic peak are 123-140 meV for S-OPE n Fc and 207-218 meV for S-CH 2 -OPE n Fc, both of which are higher than the theoretical value for a homogeneous monolayer (90 meV) [30] and the reported well-packed ferrocenylalkanethiolate SAMs. [31] In the context of the Laviron analysis, [32,33] broadening of the peaks in CV can be related to repulsive molecule-molecule interactions. Therefore, we conclude that the large FWHM values in S-CH 2 -OPE n Fc SAMs are caused by the strong repulsion between Fc groups when they are charged during oxidation. [34] The rigid OPE units give less freedom for the Fc groups to rearrange and lead to stronger repulsion compared to ferrocenylalkanethiolate SAMs. [35] This interpretation implies that the repulsion between charged Fc units is stronger in S-CH 2 -OPE n Fc than S-OPE n Fc SAMs which agrees with the more-ordered neutral SAM structures for S-CH 2 -OPE n Fc compared to S-OPE n Fc as deduced from the measured smaller tilt angles (Table 1) and the measured higher rectification ratios (associated with lower leakage currents) for S-CH 2 -OPE n Fc compared to S-OPE n Fc (see next section). ρ SAM is the surface coverage. E HOMO-CV is the HOMO energy level calculated from the anodic peak potential. The error bars represent standard deviation from three independent measurements; b) Φ SAM is the SAM work function, δE ME is the offset between HOMO and the Fermi level and E HOMO-UPS is the HOMO level with respect to vacuum. The error is 50 meV, the step size for the spectra; c) d SAM is the SAM thickness. The error is 2 Å, estimated from 10% variance in peak fitting. The value in parentheses is thickness derived from CPK models and α. ρ rel is the relative surface coverage (using S-CH 2 -OPE 2 Fc as reference); d) α is the tilt angle of the Fc-OPE plane to the surface normal. The error is 2°, estimated from the linear fitting error of peak intensities against the incident angles. E LUMO is calculated from the photon energy of the first peak in NEXAFS and the binding energy of the corresponding C 1s peak; e) E g is the optical HOMO-LUMO gap. The error is 30 meV, estimated from the step size of UV-vis spectra. E LUMO-UV is the LUMO level calculated from E g and E HOMO-UPS .

Diode Characteristics
We formed electrical contacts to the SAMs with cone-shaped tips of GaO x /EGaIn following a previously reported method. [36,37] These junctions are stable in the bias window of −1 V to +1 V with ∼84% yield of nonshorting, stable junctions (108 out of 129 junctions, Table S3, Supporting Information). Figure 3a,b shows the Gaussian mean values of log 10 |J, <log 10 |J|> G as a function of the applied bias voltage V (see Figure S21, Supporting Information for the heat maps of all J(V) data). The rectification ratios RR|J|(−1 V)/|J|(+1 V) increase with n from near unity to a maximum value of 37-50 for n = 3. These diodes with conjugated backbone perform remarkably well, with RR values of just a factor 2 smaller than the well-known molecular diode of the form Ag-S(CH 2 ) 11 Fc//Ga 2 O 3 /EGaIn (RR = 1.0 × 10 2 ) in which the HOMO is localized on the Fc unit in van der Waals contact with the top electrode and is isolated from the bottom electrode by the insulating alkyl chain. The diodes reported here perform well compared against other well-known diodes (see [ref. 38] for an extensive review), but are inferior to diodes based on two Fc units. [39] Due to this asymmetry, at positive bias, the HOMO level is outside of the bias window and the charge needs to traverse the junction via coherent tunneling. At negative bias, the HOMO of Fc can participate in charge transport and provides a hopping site for holes. [40] By replacing the insulating alkyl chain with a conjugated OPE backbone, one could argue that the molecular diode would cease to operate well due to the expected delocalization of the HOMO over the entire molecule with associated loss of asymmetry. Our DFT calculations, however, show that the HOMO remains localized on the iron atom of the Fc ensuring good performance despite the conjugated OPE backbone. The molecular diodes with n = 1 do not rectify significantly likely because the OPE backbone is too short for the charge transport mechanism to switch to hopping at the negative bias (since coherent tunneling rates increase exponentially with decreasing d). [41] Figure 3c,d shows the current decay plots as a function of the molecular length d mol (the distance between the hydrogen atom in the thiol group and the furthest carbon atom in the Fc in CPK molecular models) at V = +1.0 V and −1.0 V which we used to determine the tunneling decay coefficient β by fitting the data to where J 0 is the pre-exponential factor. Because through-bond, rather than through-space, is the dominant tunneling pathway in molecular junctions, [42][43][44][45][46] we used d mol , not d SAM , as the tunneling distance in this study. At positive bias, the HOMO does not enter the bias window and the mechanism of charge transport is off-resonant tunneling (see the energy-level diagram later in Figure 7) with β = 0.44 ± 0.02 Å −1 for S-OPE n Fc. This value of β is slightly larger than the range of previously reported values for conjugated molecular wires (0.2-0.3 Å −1 ) but lower than that for aliphatic wires (0.8 Å −1 ). Remarkably, the value of β = 0.29 ± 0.08 Å −1 for S-CH 2 -OPE n Fc SAMs is smaller than that of S-OPE n Fc, possibly due to the involvement of LUMO which will be discussed in detail later. At −1.0 V, β becomes very small for both series which indicates that incoherent tunneling dominates the mechanism of charge transport. The transition from coherent to incoherent tunneling has been observed for molecular wires of 3-4 nm length, [47,48] but in our case, the Fc unit facilitates hole injection resulting in incoherent tunneling at relatively short molecular lengths in only one direction of bias. [9,40] For these reasons, we conclude that the shortest molecules in our study operate in the coherent tunneling regime, while for molecules with n = 2 and 3, incoherent transport is important at negative bias. To shed more light on the mechanism of charge transport, we conducted temperaturedependent J(V) measurements.

Charge transport in the Marcus Inverted Region
To investigate the temperature dependence of the charge transport properties of our junctions, we stabilized EGaIn in a microfluidic channel to enable J(V) measurements in a vacuum and at low temperatures [49] (see Section S9, Supporting Information for the detailed fabrication procedure). The complete data sets recorded from all junctions are summarized in Figure S22-S27, Supporting Information. The values of J and RR of the junctions are within one standard deviation of those from the same junctions using cone-shaped EGaIn tips (Figure 3). The J(V) data for junctions with S-CH 2 -OPE n Fc and S-OPE n Fc with n = 1 are independent of T, but junctions with n = 2 and 3 are dependent on T. Interestingly, for junctions with S-CH 2 -OPE n Fc the value of E a strongly depends on the applied V while in junctions with S-OPE n Fc the value of E a is independent of V.
To elucidate this difference in the temperature dependence in more detail, we determined the value of E a by plotting the data in Arrhenius plots for each V (see Figure S22-S27, Supporting Information for all Arrhenius plots) and fitting the data to the Arrhenius equation Figure 4 shows E a plotted against V (the currents from −0.2 V to +0.2 V are lower than the detection limit of the setup and are therefore omitted). For S-OPE 1 Fc (T = 180 -270 K) and S-CH 2 -OPE 1 Fc (T = 240-300 K), the charge transport process is essentially activationless (E a ≈ 0 meV) over the entire bias window, indicating that coherent tunneling dominates the mechanism of charge transport. For S-OPE 2 Fc (E a = 18.8 ± 8.5 meV, T = 150-320 K) and S-OPE 3 Fc (E a = 88.1 ± 6.1 meV, T = 150-300 K), E a is independent of V at negative bias. This indicates incoherent tunneling in Marcus normal region is the primary charge transport mechanism in these junctions. Hence, intramolecular orbital gating does not occur in these molecules (see ref. [9]), as discussed below. For S-OPE 3 Fc the value of E a is approximately five times larger than for S-OPE 2 Fc even though both SAMs have very similar δE ME and similar structures (and therefore similar λ). This increase of E a with molecular length is perhaps even counterintuitive as, in general, long-conjugated molecules can delocalize charges better than short-conjugated molecules. We attribute this increase of E a value in S-OPE n Fc junctions to the increase in the tunnel distance d (and associated decrease in molecule-electrode coupling strength γ, see below in Table 2) resulting in a substantial decrease in coherent tunneling rates: for junctions with n = 1 coherent tunneling dominates but for junctions with n = 2 and 3 coherent tunneling rates decrease and therefore incoherent tunneling increasingly dominates the mechanism of charge transport.
For S-CH 2 -OPE 2 Fc and S-CH 2 -OPE 3 Fc, E a is close to zero at positive bias, indicating a coherent tunneling mechanism. Above +0.6 V the E a for S-CH 2 -OPE 3 Fc appears to be negative, which, upon careful observation in the Arrhenius plot ( Figure S27, Supporting Information), mainly comes from the results at >220K. This behavior has been reported for junctions with ferrocene-functionalized alkanethiolate junctions because 1) the SAMs prefer a more vertical configuration and increase the tunneling barrier width and 2) the repulsive force between ferrocenium cations leads to poor contact with the top electrode. [50] Although we cannot be certain what causes a negative E a , negative E a has been observed before in other junctions [51] which could involve re-organization of the SAM induced by electrostatic repulsion (as confirmed by cyclic voltammetry discussed earlier and observed by others52) and could explain our observations. Recently, Marcus and coworkers [53] suggested that a large change in entropy could result in negative E a . We note that the values of E a reported here are lower than those typically obtained from wet electrochemical experiments because molecular junctions lack electrolyte. Hence, the outer sphere reorganization energy is negligible and E a is dominated by the innersphere contribution. Interestingly, a bell-shaped E a versus V curve is observed at negative bias which is discussed in detail in the next section.

Numerical Modeling
Equation (3) was used to fit the E a versus V data shown in Figure 4. As shown in Figure 4a, E a in the S-OPE n Fc series is independent of the applied bias, as expected for the normal Marcus region. Further, there is no statistically significant change in E a as a function of V for positive biases in the S-CH 2 -OPE n Fc series. Therefore, the Migliore and Nitzan's model (Equations 3) was used to fit the data shown in Figure 4b for negative biases. As stated previously, this model for E a is dependent of V g , the intramolecular gating voltage. This gating voltage does not come directly from the applied bias voltage. It comes from the charge build-up in the molecule. Therefore, V g can be written as, V Here, C C * represents the capacitance and the charge, Q(V b ), on the molecule is dependent on the applied bias voltage. To represent the charging on the molecule, we used the function given by Equation (6), where A is the center of the charging function and W is the width. Figure 5 shows the fits to the behavior of the activation energy as a function of applied bias for the S-CH 2 -OPE n Fc series. For S-CH 2 -OPE 3 Fc, E a continuously increases to a maximum ≈−0.7 V and decreases at higher bias. For S-CH 2 -OPE 2 Fc, the activation energy only starts increasing after ≈−0.5 V and keeps rising to −1.0 V, while no significant voltage-dependence is observed for the S-CH 2 -OPE 1 Fc junction. The model accurately portrays the change in E a for S-CH 2 -OPE 3 Fc and S-CH 2 -OPE 2 Fc, while for the S-CH 2 -OPE 1 Fc the model shows a straight line at 0 meV, indicating a charge transport process dominated by coherent tunneling. The parameters used for these fits are listed in Table 2 and shown in Figure 6. Figure 6a shows that ΔE decreases with increasing n due to the reducing HOMO-LUMO gap with n shown in Table 1. The values of λ are close to those predicted by theory [56] (and experimentally observed excitations [57] ), but increase with n (Figure 6b), which suggests that either the shape of the parabolas becomes steeper or the reaction coordinates of the two parabolas move further away from each other, counteracting the decreasing ΔE. Other factors such as gating via applied electric field or the presence of a twist angle between Fc and OPE, or between OPE units, could also affect λ. The increasing   (Figure 6d) is consistent with the increase of η for longer molecules (Table 2). Figure 6c shows a slight decrease of C C with increasing n, which increases the effective V g . In addition, although the coupling between the Fc and the adjacent OPE changes slightly with n, the larger the aromatic group the more charge it will be able to host for a given V, resulting in an increase of Q with n. The combined effect of Q and C C gives larger V g for S-CH 2 -OPE 3 Fc than S-CH 2 -OPE 2 Fc, making the required V lower for the system to enter the MIR.
To illustrate the charge transport mechanisms in these junctions, the energy level diagrams at both positive and negative bias are shown in Figure 7. Figure 2f shows that the HOMOs of the molecules are located at the Fc groups while the LUMOs are more delocalized over the molecule. At zero bias for a given molecule, E HOMO is 0.7-0.8 eV lower than the E F , and E LUMO is more than 2 eV higher than the Fermi levels (Table 1). Figure 7a shows the energy level alignment of the junctions at +1 V bias. Neither the HOMO nor the LUMO is in the conduction window and the electrons transport by coherent tunneling. The small E a values for S-OPE 2 Fc (≈20 meV) and S-OPE 3 Fc (≈60 meV) suggest that incoherence is present in these junctions. Figure 7b shows that when −1 V bias is applied to the junction, the HOMO enters the conduction window. The Fc group can be charged due to its tendency to lose one electron to form ferrocenium (Fc + ). This charged Fc + can reduce the energy level of the LUMO to shift it inside the conduction window. The decrease in HOMO-LUMO gap makes the charge transfer between Fc and OPE more exergonic, shifting the tunneling mechanism into MIR (red parabola in Figure 1c). In S-CH 2 -OPE 2 Fc and S-CH 2 -OPE 3 Fc, the weak coupling between OPE and Au allows the electrons to localize on the LUMO. For these SAMs, the model of Migliore et al. [23] accurately predicts the bell-shaped curve of E a in Figure 4b. The intramolecular gating effect pushes the charge transport mechanism into MIR. On the contrary, the stronger coupling between molecules and the  with red curved arrows. Note, the arrows indicate electron transfer pathways; in this notation, an electron tunnels first out of the HOMO (black arrow 1) to leave behind a hole.