Determination of G, S and related PF of the molecular junctions. All the SAMs were assembled on the template stripped Au (AuTS) electrode and contacted with cone-shaped EGaIn electrode in accordance with our previous work.29 The preparation of the AuTS electrode and the SAMs were set forth in Supporting Information Section S2 and the corresponding characterizations of the SAMs were summarized in Supporting Information Section S3, including X-ray photoelectron spectroscopy (XPS) and ultraviolet photoelectron spectroscopy (UPS). It could be obtained, according to the S 2p spectra (Fig. S55 ~ S58), that all the SAMs readily formed dense monolayers on the AuTS electrode. The I 3d spectra were ascribed to the terminal I atoms substitution. Based on UPS spectra, we could see that the energy offset (δEME) between all the investigated SAMs and the bottom electrodes ranged from ~ 2.1 eV to ~ 2.4 eV and changed insignificantly, identifying that the SAMs possessed the very similar coupling strength between the SAMs and the AuTS electrode (Fig. S59 ~ 62, Table S1 ~ S4).
To analyze how substitutions by I and O atoms affected the rate of electron tunneling across molecular junctions, we measured electrical characteristic curves of the SAMs with varied molecular lengths and averaged log10|J|−V plots are shown in Fig. 2a and S63 (see Supporting Information Section S4 for details) (Here, J and V were the current density and the applied voltage). Figure S64 containes corresponding histograms of log10|J| at -0.5V. We made three important observations (Fig. 2b): 1) the current density of the SAMs declined with longer molecular length, of which dropped by ~ 4 and ~ 2 orders of magnitude for the SAMs of Cn-SH and I-Cn-SH, respectively, when varying n from 5 to 14; 2) the current density of I-(C2O)m-C2-SH only attenuated ~ 1 order of magnitude for increasing m from 1 to 4; 3) the value of tunnelling decay coefficient (β) for molecular junctions of Cn-SH (n = 5, 8, 10, 11) was 0.85 similar to previous works 30, and that of I-Cn-SH(n = 5, 8, 10, 11) and (C2O)m-C2-SH (m = 2, 3, 4) were 0.36 and 0.25, respectively, according to the Simmons equations (Fig. 2b). In addition, the conductivity (σ) (at + 0.1V) of I-C11-SH in this work was ~ 2 orders of magnitude larger than that of C11-SH, while Nijhuis reported enhancement of 3 orders of magnitude31 and Whitesides showed that halogen atoms substitution modestly affect the current density32. The value of β for molecular junctions with I-(C2O)m-C2-SH (m = 1, 2, 3, 4) was 0.18 and indistinguishable from that of conjugated molecular wires,33 indicating that the SAMs we prepared were densely packed with few defects. What’s more, the current density at -0.5V of I-(C2O)4-C2-SH was ~ 4 orders magnitude higher than that of length-matched C14-SH (Fig. 2a and S63), indicating the tunneling barrier was largely reduced by I atoms and O atoms substitutions.
Table 1
Parameters used to calculate PF for molecular junctions with Cn-SH, I-Cn-SH, (C2O)m-C2-SH, I-(C2O)m-C2- SH.
Molecule
|
l (Å) a
|
lJ(+ 0.1V)l
(A cm− 2) b
|
ε
(GVm− 1) c
|
\(\:\sigma\:\)
(µS cm− 1) d
|
S
(µV K− 1) e
|
PF
(10− 10 µW m− 1 K− 2) f
|
C5-SH
|
5.78
|
0.25
|
0.173
|
0.144
|
6.6 ± 0.5
|
6.27
|
C8-SH
|
9.02
|
0.02
|
0.110
|
0.017
|
4.8 ± 0.1
|
0.39
|
C11-SH
|
12.38
|
9.33×10− 4
|
0.081
|
1.115×10− 3
|
4.5 ± 0.3
|
2.26×10− 2
|
C14-SH
|
15.66
|
4.06×10− 5
|
0.064
|
6.343×10− 5
|
2.8 ± 0.1
|
4.97×10− 4
|
I-C5-SH
|
7.23
|
0.66
|
0.138
|
0.478
|
10.8 ± 0.4
|
55.75
|
I-C8-SH
|
10.68
|
0.11
|
0.094
|
0.120
|
9.2 ± 0.4
|
10.17
|
I-C11-SH
|
13.83
|
0.03
|
0.072
|
0.042
|
5.8 ± 0.2
|
1.41
|
I-C14-SH
|
17.20
|
8.91×10− 3
|
0.058
|
0.015
|
2.8 ± 0.1
|
0.12
|
(C2O)2-C2-SH
|
8.51
|
0.15
|
0.118
|
0.130
|
5.8 ± 0.2
|
4.37
|
(C2O)3-C2-SH
|
11.63
|
0.05
|
0.086
|
0.058
|
4.2 ± 0.1
|
1.02
|
(C2O)4-C2-SH
|
14.65
|
0.02
|
0.068
|
0.029
|
3.7 ± 0.1
|
0.40
|
I-C2O-C2-SH
|
7.08
|
24.00
|
0.141
|
17.021
|
10.8 ± 0.4
|
2.00×103
|
I-(C2O)2-C2-SH
|
10.21
|
12.88
|
0.098
|
13.143
|
11.2 ± 0.3
|
1.64×103
|
I-(C2O)3-C2-SH
|
13.26
|
4.37
|
0.075
|
5.827
|
11.1 ± 0.5
|
7.18×102
|
I-(C2O)4-C2-SH
|
16.37
|
2.10
|
0.061
|
3.423
|
11.1 ± 0.5
|
4.22×102
|
a l represents the molecular length determined by ChemDraw software considering the tilt angle (30°).
b Determined from J-V curves in Fig. 2a and S63.
c ε is the magnitude of electric field intensity and is obtained from the applied voltage across the junctions (+ 0.1 V) devided by l.
d \(\:\sigma\:\) is the conductivity of the SAMs and is obtained as \(\:\sigma\:\) = J/ε.
e The significant digit of S is 0.1 µV/K based on the accuracy of the output voltage and the accuracy of the input temperature and was detailed discussed in our previous work.13
f It is difficult to directly estimate the error bar of PF, since the values of PF are calculated from S and σ. Therefore, we utilized the average of S and σ to obtain PF of the SAMs, and we have noticed that other groups also only reported the error of S and σ but not PF.7,8,10,34
To characterize S of molecular junctions, we recorded potential difference (ΔV) in response of temperature difference (ΔT = 2.0 K, 3.5 K, 5.0 K) applied on the SAMs with heating the bottom AuTS electrode and cooling the top EGaIn electrode. The corresponding setup and analysis protocol of thermoelectric measurements was reported by our previous study as shown in Fig. S65 (see Supporting Information Section S5 for details).13 Here, we give a brief description. For each SAM at applied ΔT, we measured at least 20 junctions to generate statistical significant data to determine average of ΔV (<ΔV>) and related standard-deviation by Gaussians fitting (Fig. S66 ~ 70). Then, the values of S for molecular junctions can be obtained by plotting < ΔV > as a function of ΔT (Fig. 2c and S71). After detailed analysis of dependence between measured < ΔV > and S (reported in our previous work), we acquired the values of S for the SAMs as listed in Table 1. We observed: 1) the positive sign of S for the investigated molecules indicated HOMO dominated charge transport properties in all cases; 2) engineering molecules by terminal I atom substitution and replacing backbone -CH2- unite with O atom could result in the enhanced values of S comparing with Cn-SH; 3) S for the SAMs of Cn-SH, I-Cn-SH and (C2O)m-C2-SH declined with increased molecular length; 4) the values of S for the SAMs of I-(C2O)m-C2-SH, ~ 11 µV/K, were nearly independent on molecular length, and that of I-(C2O)4-C2-SH was ~ 4 times higher than C14-SH (Fig. 2d). Therefore, combining these substitutions could enhance the Seebeck coefficient of the SAMs, which is a crucial step toward enhancing the thermoelectric performance of molecular junctions.
To evaluate the enhancement of the thermoelectric effect, we calculated the values of PF of each SAM (Fig. 2e and Table 1) and the detailed processes were reported in Supporting Information Section S5. It is interesting that values of log10PF linearly attenuated with the molecular length, which has not been reported before. We think the reasons could be 1) terminal I atom substitution and replacing backbone methylene units (-CH2-) by O atom improved the current density ranging from 2 to 4 orders of magnitude and the measured log10|J| followed a good linear correlation with molecular length, 2) the increase of S by the atomically tuned chemical structures was less than 5 times and the sharply promoted PF was mainly originated for the change of G, resulting in the calculated PF being basically linearly related to the molecular length. The values of PF enhanced by 1–2 orders of magnitude through I atoms substitution, so does backbone O atom substitution. When m = 1 for I-(C2O)m-C2-SH, its PF improved by ~ 3 orders of magnitude compared with C5-SH. The longer the molecular length, the greater of degree of improvement. The PF of I-(C2O)4-C2-SH was ~ 8.5\(\:\text{×}\)105 times higher that of C14-SH, demonstrating our molecular design could dramatically boosted PF of the SAMs. As reported by Agraït19 and Lambert35, the occurrence of resonant states in τ(E) near EF could bring enhancement both for S and G. Therefore, we proposed that the molecular engineering in this work not only increased SAMs/EGaIn electrode coupling, reducing barrier of electron tunneling, but also modified mechanism of charge transport from coherent tunneling to superexchange tunneling via lone-pair electrons on O atoms. Under synergistic effect of these substitutions, the molecular energy levels of I-(C2O)m-C2-SH resonated with EF of electrodes, which resulted in the simultaneously enhanced S and G and approximately 6 orders of magnitude improvement of PF for I-(C2O)4-C2-SH compared with C14-SH.
Theoretical estimation of G, S and PF of the molecular junctions. To gain a deeper insight into the above mentioned results that atomically precise engineering of the organic molecules could improve S and G, we carried out theoretical transport calculations for the molecular junctions with Cn-SH and I-(C2O)m-C2-SH as shown in Fig. 3. The electronic structure calculations were performed at the DFT level using SIESTA36. The transmission spectra of molecular junctions were calculated using the non-equilibrium Green’s function (NEGF) method implemented in TRANSIESTA37,38. Afterwards, the linear response scheme was adopted to obtain the thermoelectric coefficients.39–42 The G, S and PF can be written as follows:
$$\:\:\:\:\:\:\:\:G(\mu\:,\:T)={\left.\frac{I}{{\Delta\:}V}\right|}_{\varDelta\:T=0}={e}^{2}{L}_{0}$$
1
$$\:S(\mu\:,\:T)={\left.-\frac{{\Delta\:}V}{{\Delta\:}T}\right|}_{I=0}=-\frac{1}{eT}\frac{{L}_{1}}{{L}_{0}}$$
2
$$\:\text{P}\text{F}=G{S}^{2}$$
3
Here the coefficient Ln is dependent on the transmission coefficient.
$$\:{L}_{n}(\mu\:,\:T)=\frac{1}{h}{\int\:}_{-\infty\:}^{\infty\:}{\left(E-\mu\:\right)}^{n}T\left(E\right)(-\frac{\partial\:f(E,\mu\:,T)}{\partial\:E})dE$$
4
The optimized structure of molecular junctions of Cn-SH and I-(C2O)m-C2-SH are shown in Fig. 3a and S72 ~ S73, where we utilized Au as the top electrode and use a constant self-energy to avoid the complication in modeling the EGaIn electrode, which has been done previously43–45. The molecule together with one extra layer of Au from each side were fully relaxed until the forces acting on each atoms were less than 0.02 eV/Å.
The transmission coefficient of Cn-SH gets lower as the molecular length increases as shown in Fig. 3b and S74, which is the consequence of broader tunneling barriers for longer molecule. New resonances in the junctions formed by I-(C2O)m-C2-SH are found around E = -1.83 eV (as indicated by the black dashed line in Fig. 3c). Since they (E = -1.83 eV) are further away from the calculated Fermi energy (E = 0 eV), their contribution to transport properties is negligible. However, in previous experimental studies signature of enhanced conductance due to these states were reported44. Several reasons may results in such discrepancy between theoretical and experimental results. One is that the standard DFT calculation is not accurate enough, especially in determining the Fermi level. The second could be that, the coupling between different molecules may lead to broadening these resonant states. To understand qualitatively the experimental results, we have chosen to take the Fermi energy as a tuning parameter. We have calculated the thermoelectric transport coefficients (G, S and PF) using different Fermi energy values for Cn-SH and I-(C2O)m-C2-SH, which are shown in Fig. S75 ~ S77. We find that only when the Fermi level is located near these resonant states, could the theoretical results be consistent with experimental ones (Fig. S78 ~ S79).
We have shown theoretical results for EF = -1.83 eV in Fig. 3. In this case, the improvement of both G and S for I-(C2O)m-C2-SH observed in the experiments can be reproduced from the DFT calculations (Fig. 3d, e). We have the following observations (Fig. 3f): 1) the values of PF for the SAMs of Cn-SH gradually declined with the increased molecular length; 2) the PF of I-(C2O)m-SH were always higher than those of Cn-SH for the equivalent molecular lengths; 3) the PF of I-(C2O)4-C2-SH was approximately 4 orders of magnitude larger than C14-SH. Therefore, the theoretical results offered a qualitative understanding of the experimental measurements, indicating that the molecular engineering method in this work could significantly increase the PF of molecular junctions, which is promising for further development of organic thermoelectric materials at nanoscale. In addition, the measured S of the SAMs of I-(C2O)m-C2-SH derivatives were much lower than the theoretically predicted values, highlighting that futher efforts should be devoted on the finely-tuning the electronic structures of molecular junctions to engineering the quantum transport properties for the high-performance thermoelectronic devices.
In conclusion, we focused on how the heteroatom substitution, including the terminal I atom substitution and replacing the backbone -CH2- by O atoms, affected the charge transport across molecular junctions, and confirmed that both G and S could be improved in the SAMs of I-(C2O)m-C2-SH (m = 1, 2, 3, 4), compared to the length-matched Cn-SH (n = 5, 8, 11, 14). According to the electrical tunneling results, the values of β for the SAMs of Cn-SH and I-(C2O)m-C2-SH were 0.85 and 0.18, respectively, indicating the sharply declined tunneling barrier, and the value of σ (at + 0.1V) for I-(C2O)4-C2-SH was more than 4 orders of magnitude larger than that of C14-SH. In addition, the values of S for the SAMs of Cn-SH, I-Cn-SH and (C2O)m-SH declined with increased molecular length, while that for the SAMs of I-(C2O)m-C2-SH, ~ 11 µV/K, were nearly independent on molecular length. The S of I-(C2O)4-C2-SH was ~ 4 times higher than that of C14-SH. Therefore, the PF of I-(C2O)4-C2-SH was dramatically enhanced by a factor of 8.5\(\:\text{×}\)105 compering with C14-SH. Based on the DFT calculations, the underline mechanism was that the new resonances were introduced into the SAMs after heteroatom substitution near the Fermi energy. This work highlighted the importance of molecular engineering for nano thermoelectric devices with high-efficiency conversion of the wasted heat to the useful electricity.