Intermolecular and Electrode-Molecule Bonding in a Single Dimer Junction of Naphthalenethiol as Revealed by Surface-Enhanced Raman Scattering Combined with Transport Measurements

Electron transport through noncovalent interaction is of fundamental and practical importance in nanomaterials and nanodevices. Recent single-molecule studies employing single-molecule junctions have revealed unique electron transport properties through noncovalent interactions, especially those through a π–π interaction. However, the relationship between the junction structure and electron transport remains elusive due to the insufficient knowledge of geometric structures. In this article, we employ surface-enhanced Raman scattering (SERS) synchronized with current–voltage (I–V) measurements to characterize the junction structure, together with the transport properties, of a single dimer and monomer junction of naphthalenethiol, the former of which was formed by the intermolecular π–π interaction. The correlation analysis of the vibrational energy and electrical conductance enables identifying the intermolecular and molecule–electrode interactions in these molecular junctions and, consequently, addressing the transport properties exclusively associated with the π–π interaction. In addition, the analysis achieved discrimination of the interaction between the NT molecule and the Au electrode of the junction, i.e., Au−π interactions through-π coupling and though-space coupling. The power density spectra support the noncovalent character at the interfaces in the molecular junctions. These results demonstrate that the simultaneous SERS and I–V technique provides a unique means for the structural and electrical investigation of noncovalent interactions.


S2. Scaling of flicker noise in a single-molecule junction
Before discussion of conductance fluctuation of molecular junction, we mention the expression of the conductance of the junction. The conductance (G) can be derived from the Landauer formular with Breit-Wigner resonance model, as shown equation S1. 4 , where E represents the energy of the electron, ε０ represents the molecular orbital energy with respect to the Fermi energy, G0 represents the quantized unit of electrical conductance (2e 2 /h) and Γ represents the broadening of the molecular orbital induced by the metal-molecule interaction. Then, the fluctuation of the conductance (<ΔG 2 >) can be obtained by the derivative of (S1).
Assuming that the Γ derived from the interaction was expressed by the product of exponential term and the fluctuation term (p(z)).
, where β represents the decay constant. On the condition that fluctuation of p(z) can be expressed by the Gauss function, the fluctuation of Γ can be expressed as follows. 1 By substituting eq. (S4) to eq. (S2) gives eq. (S5) The conductance fluctuates depend on the GAVE 2 . When we assume the dimer junction, we regard the electron transport via the levels that is correlated with the interaction term (δ). Similar to eq.1, the conductance of the junction can be expressed as shown in eq. (S6), assuming the δ is much smaller than E and ε０. The derivative of (S6) gives Therefore, Thus, the conductance fluctuation <ΔG 2 > depends on the (Δδ) 2 GAVE 2 and (ΔΓ) 2 GAVE. In the case that the fluctuation of the interaction between molecules mainly contributes to conductance fluctuation, <ΔG 2 >∝GAVE 2 . On the other hand, the case that the fluctuation of the metal-molecule interaction mainly contributes to conductance fluctuation, the <ΔG 2 >∝GAVE. Even in the dimer junction, the both molecule-molecule and metal-molecule interaction can be contributed to the conductance fluctuation, the conductance fluctuation have the value of 1< <ΔG 2 > <2. [5][6] We then estimated the scaling of the noise power following the previous method for H and L states. 1 We calculated the correlation factor between Sn and G λ regarding the scaling exponent (λ) as shown in Figure S2. The scaling exponent of the system was determined as the value makes the correlation coefficient zero. Before discussing S-4 the scaling exponent, we have to judge the validity of the data set for the correlation analysis. While even the maximum correlation factor was 0.4 in the H state, that was only 0.1 in the low states. Namely, Sn and G are weakly correlated in the L state, while those show no correlation in the H state. It can be considered that the small conductance window for the analysis in the H state derived from the significant modulation in the binding energy of Au-π interaction hampers the precise estimation of the scaling.
According to previous studies, polycyclic aromatic hydrocarbons such as benzene and naphthalene physisorbed on the Au surface show a variation in the binding energy from 0.5 eV to 1.5 eV. [7][8][9][10] Under the room temperature measurement of around 300 K, the molecule and atoms are thermally fluctuated among these states and modulate the Au-Naphthalenethiol (NT) distance, prompting immediate geometrical changes from the H state to the L state. The immediate change from the H state to the L state in the Au-π interaction is validated by the separation dependence of the conductance and as shown in Figure 4. The conductance significantly changes in response to the separation distance. The immediate change from the rigid bond with high conductivity to the weak bond with low conductivity narrows the distribution in the conductance in the H state and broadens that in the L state. Figure S2. Correlation of the noise power and condutance for H and L staete.

S3. Geometry of the single-molecule junction of the theoretical simulation.
We built the junction structure and calculated the transmission spectra using the Virtual NanoLab -Atomistix Tool Kit package (Quantum ATK ver. 2014.2), which is based on the DFT combined with the nonequilibrium Green's function method. 11 The optimization and transport calculation were performed by the Perdew-Burke-Ernzerhof (PBE) generalized gradient approximation (GGA) exchange-correlation energy functional. The electrode was built from the (111) direction of the periodic gold structure in the ATK package. The top layer consists of three gold atoms for the connection to NT. Thiol atoms were initially set on the hollow site and the geometry was optimized until the maxim force on each atom is less than 0.05 eV/Å. A cutoff energy of 75 Hartree with the 3x3x50 k-point mesh were applied. The basis functions were Single-Zeta polarized for Au, and Double-Zeta polarized for the other atoms. After the optimization, the S-Au distance is 2.5 Å and distance between two naphthalene rings were 3.1-3.4 Å, agreeing with the previous study. 10,12 To evaluate the electron transport through the NT junction, we constructed the models with a variety of separation distances to find the preferable junction structure in the breaking process of the junction. The number of the models was up to 33 and 58 models for monomer and dimer junctions, respectively.
The optimized structure and calculated transmission spectra at the separation were shown in Figures S3 and S4 for Au/NT monomer/Au junction and Au/NT dimer/Au junction.
We, then, rebuilt the junction structure for the vibrational calculation from the optimized structure. We construct the electrode with three gold layers. The structure converted into the Gaussian 16 format, and the junction geometry of the junction was optimized by the density functional theory (DFT) method of CAM-B3LYP with Grimme's D3BJ empirical dispersion correction. 13 The basis function were LanL2DZ for gold, and 6-31G(d) for carbon, hydrogen, and sulfur atoms to include the effect of the dispersion force. The scaling factor was 0.95. 14 Figure S3. Transmission spectra depending on the separation distance for Au/NT/Au junction. Figure S4. Transmission spectra depending on the separation distance for Au/NT dimer/Au junction.

S4. Two-dimensional conductance histogram.
We confirmed the formation of the NT junction based on the mechanically-controllable break-junction method. [15][16][17] Figure S5 (a) shows typical conductance traces measured with clean Au (black line) and NT-modified (blue lines) electrodes. In the case of the clean Au electrode, a single distribution appeared at 1 G0 in the 2D histogram ( Figure   S5(b), left) complied from the conductance traces, which arises from the Au atomic junction. In contrast, the 2D histogram exhibited cloud-like distributions in the range of 10 -3.5 -10 -1.5 G0 in the presence of the NT molecules on the electrode ( Figure S5(b), right), and these distributions demonstrate the formation of NT molecular junctions.
Importantly, the conductance range for the NT molecular junctions observed in the MCBJ study agrees well with the one found in the I-V measurements simultaneously acquired with SERS (Figure 2(c)). Thus, the results in Figure   S5 prove the formation of the NT molecular junctions in the simultaneous SERS and I-V measurements discussed in the main text.
It is noteworthy that the conductance states of the NT molecular junctions appeared as cloud-like distributions in the 2D histogram ( Figure S5(b), right). We ascribed this to insufficient mechanical stability of the NT dimer junction against the external force exerted during the MCBJ process. The binding energy of the NT dimer was estimated to be around 0.35 eV (see Section S7), while in the case of terphenylenedithiol dimer, which afforded distinct distribution in the 2D conductance histogram, the binding energy of approximately 2 eV was found. 6 Despite the blurred appearance of the NT molecular junctions in the 2D conductance histogram, their conductive states were clearly resolved in the 2D mapping of the conductance and the Raman shift (Figure 3(a)). This difference arises most probably due to the absence of the external force in the simultaneous SERS and I-V measurements, which involves no mechanical displacement of the electrode. S-8 Figure S6 represents SERS spectra of Au/NT dimer/Au junction, histogram of the vibrational energy, Raman spectra of the powder sample and the calculated spectra of Au/NT dimer/Au junction. The SERS spectra qualitatively resembles with the powder sample and theoretical simulation, allowing to the vibrational mode observed around 830 cm -1 , 1070, 1380, and 1620 cm -1 , were assigned to the vibrational mode marked (1), (3), (6), and (10) in the calculated spectrum. The displacement of the atoms for each vibrational mode were displayed in Figure S7. We show the example of the SERS spectra and corresponding I-V curves for the high and low conductive states, respectively, in Figure S8.

S6. SERS signal intensity at the molecular junction.
In the single-molecular junction, SERS signal was enhanced by the electromagnetic field effect and charge transfer effect, though the electromagnetic field effect also affects the molecule at the relevant of the junction on the gold surface, the molecule bridging over metal nanogap obtains additional charge transfer effect, leading to the strong SERS signal in the molecular junction. [18][19][20][21] Figure S9 illustrates pairs of the simultaneously measured SERS spectrum and I-V curve obtained with the NT-modified electrodes. For Figure S9(a), the conductance was determined to be 10 -3 G0 based on the I-V curve. This value agrees with the conductance of the NT single-molecule junction (Section S4), demonstrating the presence of the molecular junction. In contrast, for Figure S9(b), the conductance was as low as 10 -6 G0 significantly lower than the single-molecule conductance of NT, and the molecular junction was absent during the measurements. Comparison between the SERS spectra in Figure S9 reveals that the SERS intensity of the molecular junction was significantly enhanced compared with that of disconnected states. Figure S10 shows the relationship between the SERS intensity and the conductance. It was confirmed that the SERS intensity significantly increased. We also observed the blinking of the SERS signal depending on the conductance value and the intensity of the ring breathing mode was enhanced in the region of 10 -3.8 -10 -2 G0 corresponding to the conductance of the molecular junction ( Figure S10

S7. Relationship between vibrational energy and conductance.
The relationship between the vibrational energy of C-H bending mode (νC-H bending) and conductance in a log scale (log (G/G0)) was depicted in Figure S11. The L1 and L2 states were clearly separated into two distinctive regions.
The positive correlation in the vibrational energy between ring breathing mode (νring breathing) and C-H bending mode supports the consistency of the spectroscopic observation. Some states, especially in the high conductive region, cannot be observed, because the peak intensity is different among the vibrational and small intensity is under the detection limit.

S8. Displacement dependence of Energy for NT dimer
We calculated the energy of the NT dimer depending on the displacement along the long axis of the dimer. The geometric optimization was performed in Gaussian 16 format by the method of CAM-B3LYP with Grimme's D3BJ empirical dispersion correction. The basis function was 6-311++G(d,p). We calculated the energy of the dimer depending on the displacement by shifting the top monomer from 0 to 8 Å, with an interval step of 0.4 Å with fixing the S atoms ( Figure S8). The binding energy of the dimer was estimated to be 350 meV by subtracting the energy at the separated state from the that fully overlapped state. The estimated binding energy agrees well with the previous study. 10,22 Figure S12. Displacement dependence of energy of NT dimer. Inset represents the optimized structure at the corresponding-colored circle.

S9. Distribution of the vibrational energy focusing on the Au/NT dimer/Au junction.
The two-dimensional histogram of conductance and Raman shift of ring breathing mode focusing on the L-state was depicted in Figure S13. Even lower conductance region than the L2 state, the distribution of the vibrational mode contains at least two components whose peak centers are 1066 cm -1 and 1067 cm -1 . The smaller vibrational energy might originate from dimer junction and lager one might originate from disconnected states. Figure S13. Two-demensional histogram of condutance and Raman shift.

S10. Evolution of the NT monomer junction.
DFT calculations were performed to investigate the evolution of the NT monomer junction as the increase in the gap width between the electrodes ( Figure S15). The computational details were the same as those employed for the NT dimer (Section S8). The total energy of the NT monomer junction was found to monotonically increase as the gap width increases. The slope of the energetic changes around the junction structure with the lowest energy is 174 meV/Å. This value is substantially larger than the corresponding value found for the NT dimer (48 meV/Å, see Section S8). The steep energetic increase indicates that the monomer junction adopts very limited structure leading to the very limited variation in the junction conductance. Figure S15. Evolution of the energy of the NT monomer junction calculated by DFT.