High thermopower of mechanically stretched single-molecule junctions

Metal-molecule-metal junction is a promising candidate for thermoelectric applications that utilizes quantum confinement effects in the chemically defined zero-dimensional atomic structure to achieve enhanced dimensionless figure of merit ZT. A key issue in this new class of thermoelectric nanomaterials is to clarify the sensitivity of thermoelectricity on the molecular junction configurations. Here we report simultaneous measurements of the thermoelectric voltage and conductance on Au-1,4-benzenedithiol (BDT)-Au junctions mechanically-stretched in-situ at sub-nanoscale. We obtained the average single-molecule conductance and thermopower of 0.01 G0 and 15 μV/K, respectively, suggesting charge transport through the highest occupied molecular orbital. Meanwhile, we found the single-molecule thermoelectric transport properties extremely-sensitive to the BDT bridge configurations, whereby manifesting the importance to design the electrode-molecule contact motifs for optimizing the thermoelectric performance of molecular junctions.


Repeated formations of Au-BDT-Au molecular junctions
Au-1,4-benzenedithiol (BDT)-Au junctions were formed by using a microheaterembedded mechanically-controllable break junction (MCBJ). In this device, a freestanding Au nanocontact was mechanically broken through the substrate deflection by a three-point bending mechanism (Fig. S1a). Here, we used a piezo-actuator to push the MCBJ beam from the back side that moves in a vertical direction by the applied dc voltage Vpiezo at a rate 1 μm/V. Furthermore, the device configuration was designed to provide the attenuation factor r = ηdj / Dpiezo = 3 × 10 -4 , where dj is the tensile displacement of the Au junction S1,S2 induced by the beam bending through the motion of the piezoactuator by a distance Dpiezo and η is a coefficient describing the mechanical deformation of the polyimide layer. S3 This enables fine control of the contact mechanics at a subpicometer resolution thereby allowing reproducible formations of stable atomic and molecular junctions. S4 In prior to the single-molecule measurements, the MCBJs were immersed in a dilute toluene solution of BDT (1 μM). The junction was then broken at the narrowest constriction through the substrate bending to let BDT molecules adsorb on the fresh Au S3 breakage surface exposed to the solution via Au-S links. The chamber was then evacuated to remove the solvent for preventing molecular aggregation on the junction. S5 In experiments, the Pt microheater adjacent to the Au nanobridge ( Fig. S1b) was heated by applying a dc voltage Vh and the MCBJ substrate was bent at varying speeds at above 1 nm/s depending on the conductance states G during junction elongation until the Au contact was narrowed mechanically to a size of a few atoms with G below 6 G0.
Thereafter, on the other hand, the junction stretching speed was set to 6 pm/s. Under this slow stretching speed condition, the Au atomic contacts underwent thermoactivated spontaneous breakdown showing long 1 G0 plateaus in G -t traces that last for longer than 10 seconds. S6 After the junction fracture, BDT molecules were often bridged the thus formed two Au nanoprobes, which was observed as another plateaus at a low-G regime in a range from 0.1 G0 to 0.001 G0 (Fig. S1c). Meanwhile, we recorded the thermoelectric voltage and the conductance of the junctions when G decreased below 8 G0 (Fig. S1d). Further stretching, G dropped to below 10 -4 G0 indicating BDT molecular junction breakdown at the Au-S bonds. Subsequently, the junction was closed at a rate 1 nm/s until G exceeded 15 G0. The entire processes were repeated for 50 times at each

S4
Vh condition from 1.0 V to 5.0 V. All the measurements were conducted at room temperatures in a vacuum better than 10 -5 Torr. Only a part of the data is shown for the second trace (black) for the sake of clarity.

Simultaneous measurements of thermoelectric voltage and conductance
The thermoelectric voltage at the atomic and molecular wires ΔV were acquired by measuring the potential drop ΔVc at the 100 kΩ sensing resistor connected in series to the junction together with the conductance G (Fig. S2a). S7 Briefly, we first recorded G under the dc voltage Vb = 0.2 V applied to the junction when G < 8 G0 in course of the aforementioned junction stretching processes. Subsequently, we switched off the voltage source and measured ΔVc. The sequential recording was performed until G decreased from 8 G0 to 0.0001 G0. A constant bias voltage was imposed to the microheater throughout the measurement to create a temperature gradient at the junction for inducing measurable amount of ΔVc. Because of the long integration time required for measuring the small thermoelectric voltage at the high-resistance molecular junction with accuracy, the sampling rate of G and ΔVc was as slow as 3 Hz.
The measured thermoelectric voltage was found to include a background presumably stemming from the thermoelectric effects at the bulk interconnects. S7 This background voltage was calibrated by acquiring ΔVc at Vh = 0 V for 10 junction formation/breaking processes at every each Vh condition measured (red plots in Fig. S2b for example). By subtracting the thus obtained background from ΔVc, we deduced the actual thermoelectric S7 voltage occurring at the junction by taking into account the voltage division in the circuit as ΔV = ΔVc(1 + 10 -5 /G) (Fig. S2c). S7

Temperature distributions in a microheater-embedded MCBJ
Heat transfer in the microheater-embedded MCBJs were simulated theoretically by COMSOL. The three-dimensional structure was defined following the actual dimension and material used: A polyimide layer of thickness 4 μm was put on a phosphor bronze substrate; on the polyimide, 40 nm thick Al2O3 layers were drawn; and on the top lies a Au junction of thickness 100 nm with a platinum microheater adjacent to its left side at a distance of 300 nm. The acute angles at the tips of the Au junction were set to 45 • and the narrowest constriction was defined to be 50 nm to 1 nm wide. In the simulation, we set the temperature of the microheater at Th = 500 K while assuming the bottom plane of the 5 μm phosphor-bronze layer at room temperature Tc = 293 K. A heat transfer in solids module of COMSOL has been employed to estimate the temperature distribution, and the simulation results are displayed in Fig. S3.
The results show a substantial temperature drop at the junction due to the geometrical constriction. It is also noticeable in Fig. S3c that heat leakage through the substrate is sufficiently suppressed in a direction along the junction due to the free-standing structure together with the micrometer-scale deep etching of the polyimide. Figure S3e illustrates the temperature distribution along the axis pointing from higher temperature to the lower S10 one for junctions having a constriction of size 50 nm, 10 nm, or 1 nm (We note that the present simulation is not valid for 1 nm contacts as it calls for a theory that fully takes quantum effects into account; we show the result here just to show the qualitative tendency). It gives an average temperature gradient of 3.2 K/nm at the junction having a contact of size 50 nm × 50 nm. Although the actual temperature profile cannot be obtained for molecular junctions, these results ensure that substantial amount of temperature difference would be established at the atomic and molecular bridges considering their low thermal conductance compared to the large Au contacts.
While the thermal analysis reveals the capability of the microheater to create a Kelvins of temperature difference at the junction, Fig. S3 also indicates that the temperature drops largely in the micro-scale Au lead. This makes the thermovoltage generated in the Au lead to be canceled there as described in the previous work. S7

Figure S3. Heat transfer analysis for a microheater-embedded MCBJ. a,
A three-dimensional model of a microheater-embedded MCBJ having a 1 nmsized Au contact and the temperature distributions around the junction in case when the temperature Th at a Pt microheater was set to 500 K. b-c, Top (b) and side views (c) of the temperature distributions at Th = 500 K. d, Threedimensional model showing a magnified view of a Au junction in a microheater-S12 embedded MCBJ. The contact size is narrowed to 50 nm scale in the image. e, Temperature profile along the junction with a contact of size 50 nm x 50 nm (blue), 10 nm x 10 nm (red), and 1 nm x 1 nm (yellow) modeled at the middle.
Note that the temperature at the heat downstream tends to be closer to 293 K as the contact size become smaller. The inset shows a magnified image of the junction with dotted lines denoting the 500 nm-long constriction. The position of the middle part of the junction was taken to be x = 0.   Figure S7. Conductance versus thermoelectric voltage two-dimensional histograms at various Vh conditions. Positive and negative ΔV is found at G S17 > 1 G0 and G < 0.1 G0, which are ascribed to thermoelectricity in Au atomic wire and BDT molecular junctions, respectively. The thermovoltage is rising steadily with increasing Vh suggesting larger temperature gradient under higher Vh. S18 Figure S8. Average thermoelectric voltage. The average thermoelectric voltage ΔVave plotted against Log10G.