SUPPORTING INFORMATION Promoting Dual Electronic and Ionic Transport in PEDOT by Embedding Carbon Nanotubes for Large Thermoelectric Responses Kyungwho Choi, 1,2+ Suk Lae Kim, 1+ Su-in Yi, 1 Jui-Hung Hsu, 3 Choongho Yu 1,3* 1 Department of Mechanical Engineering Texas A&M University College Station, Texas 77843 USA 2 New Transportation Systems Research Center Korea Railroad Research Institute Uiwang-si, Gyeonggi-do, 16105 Korea 3 Department of Materials Science and Engineering Texas A&M University College Station, Texas 77843 USA + Contributed equally to this work. * Corresponding author. Email: chyu@tamu.edu 1
1. Electrical Property Measurement As-prepared samples on glass slide substrates (width: 7~8 mm, length: 25 mm) were mounted between two commercial Peltier devices (Marlow industries). To create a temperature gradient along the samples, two Peltier devices were used as shown in Fig. S1. When the samplefacing side for one Peltier device is heated, that of the other Peltier device is cooled. Temperature of each Peltier device was varied between +4 K and -4 K to create a maximum temperature difference of 6~8 K. T-type thermocouples that consist of copper and constantan wires placed on the sample to measure the temperature of hot and cold side of the sample. To minimize the electrical contact resistance, metal electrodes were made on the sample with silver paint. Fourprobe electrical resistance measurement was conducted by making two additional metal electrodes at both sides of the samples. Ohmic contacts between the thermocouple and samples were confirmed by linear current-voltage relations (Fig. S2a). To obtain the thermopower of samples, 6~8 different temperature differences were created and corresponding thermoelectric voltages were recorded. We ensure identical locations for the temperature and thermoelectric voltage measurements by measuring voltage using the copper wire of the thermocouple. The thermopower of samples was calculated from the linear slope, as shown in Fig. S2b. Voltage measurement with copper wires of the thermocouples T-type thermocouple T-type thermocouple Electrode Peltier device Peltier device T= -10 ~ +10 K Aluminum heat sink 25 mm 2 mm11~12 mm <Top view> 7~8 mm Thin film sample (80~110 nm) Glass slide <Side view> Figure S1. Schematic of a setup for measuring electrical conductivity and thermopower. 2
(a) Voltage (V) 15 10 5 0-5 -10-15 -8-6 -4-2 0 2 4 6 8 Current ( A) Figure S2. Representative measurement data of Sample L after 30-min TDAE exposure. (a) I-V curve of the Sample L, showing the Ohmic contact between sample and measurement electrodes. (b) Thermoelectric voltage from Sample L as a function of temperature difference between the heating and cooling side. (b) Thermoelectric voltage (V) 0.12 0.08 0.04 0.00-0.04-0.08-0.12-8 -6-4 -2 0 2 4 6 8 Temperature difference (K) For the all samples after the reduction process, we first found a stable state by measuring electrical conductivity periodically, and then we collected actual data displayed in the main manuscript. For example, Fig. S3 shows the electrical properties of Sample L after 30-min reduction as a function of time in air environment. After ~30 hours, electrical conductivity of the reduced sample became steady, and thermopower was also kept constant. Figure S3. Electrical properties of Sample L after 30-min reduction as a function of time. Actual data was collected after the electrical conductivity became stable. 3
2. Carrier Concentration and Mobility Hall measurements with a Van der Pauw geometry were performed under 1-T magnetic field with a custom-built setup following the ASTM F76-08 method. 1 The property changes upon the variations of TDAE exposure time were further studied by measuring the hole carrier concentration and mobility of Sample L at 300 K (Fig. S4a) by using the Hall measurement technique with the Van der Pauw geometry. Prior to the TDAE exposure, the carrier concentration was ~7.0 10 20 cm -3, comparable to that of our PEDOT-Tos only sample (Fig. S4b). After 10-min reduction, the carrier concentration was remarkably dropped to ~3.8 10 17 cm -3, and then it was saturated to ~1.0 10 17 cm -3 with the longer exposure. The mobility (~0.81 cm 2 /V-s) before the reduction was largely improved to ~6.1 cm 2 /V-s after a 10-min exposure, and reached to ~8.0 cm 2 /V-s with longer exposure. It should be noted that the numbers here are given to have qualitative pictures about the change of carrier concentration and mobility as a function of the TDAE exposure time. It is difficult to quantitatively discuss the mobility and hole concentration because the thermoelectric effect is caused by combination of ion and electron transport. The ion mobility is significantly lower than that of electron, so the ion effect on the mobility might be relatively small. Figure S4. Hole carrier concentration (n) and mobility (µ) of Sample L and H. (a) n and µ of Sample L vs. TDAE exposure time. n and µ were saturated after 30-min reduction. (b) n and µ of PEDOT-Tos only, Sample L, Sample H, and CNT only before (solid lines) the TDAE exposure and after (dashed lines) 30-min TDAE exposure. Sample L and H contain CNTs whose mobility is higher than that of PEDOT-Tos, but the mobility of Sample L and H was kept almost unchanged from that of PEDOT-Tos before the TDAE reduction. This relatively low mobility was significantly raised after the TDAE exposure. 4
The role of CNTs was elucidated in Fig. S4b before and after reduction by comparing the carrier concentration and mobility of Samples L and H along with PEDOT-Tos only and CNT only samples. Prior to the reduction, the low mobility (~0.75 cm 2 /V-s) of PEDOT-Tos, which is slightly higher than that of popular undoped (i.e., no additives) PEDOT:polystyrene sulfonate, 2 was marginally increased (~0.81 cm 2 /V-s) by adding CNTs whose mobility is relatively high (~6.3 cm 2 /V-s). However, after the reduction, the mobility of Sample L was remarkably improved by a factor of 10 while that of CNTs became only twice higher. 3. Thermal Conductivity Measurement The thermal conductivity of Sample L and PEDOT-Tos was measured by using the microdevice shown in Fig. 4a. Samples prepared on a substrate were detached and placed between two suspended membranes of the microdevice. In the microdevice, each membrane is thermally isolated by 6 SiNx beams, and the membrane has a serpentine platinum resistor thermometer (PRT) for heating and temperature measurements. When a DC current is supplied through the PRT of a membrane (heating membrane), the heat generated by Joule heating is transferred to the other membrane (sensing membrane) through the bridged sample. The thermal conductance of the sample, Gs can be calculated as: G T s s Gb (1) Th Ts where Gb is the thermal conductance of the beam and ΔTs and ΔTh are temperature increase with respect to the environmental temperature (T0) in sensing membrane and heating membrane, respectively. ΔTs and ΔTh can be obtained from the PRT in each membrane, and Gb can be obtained as: G Q Q h b b (2) Th Ts where Qh and Qb are Joule heating power caused by the PRT in the heating membrane and the platinum pattern on the beam, respectively. Then geometrical parameters are multiplied to get the thermal conductivity (k) of the sample: l (3) k Gs tw 5
where l is the gap distance between the two membranes, t is the thickness of the sample, and w is the width of the bridged portion of the sample. Figure S5. Thermal conductivity of Sample L before the TDAE exposure. See Fig. 4b (Sample L after the TDAE exposure) for comparison. We observed that thermal conductivity before and after the TDAE exposure did not make noticeable differences. We believe this is because the electronic contribution to the total thermal conductivity is small due to the relatively low electrical conductivity of our samples. 4. Monte Carlo (MC) Simulations MC simulation is often employed for explaining various physical phenomena with random nature 3. As the samples in our work are made by spraying CNT solutions, we expect that most CNTs formed random networks. Therefore, the location and orientation of CNTs were determined by using random numbers generated by the MC technique. The randomly distributed CNT model was used for calculating the thermal conductivity of the hybrids. For the simulation, we assumed that the CNTs are straight with a fixed diameter and length. The averaged length and diameter of CNTs obtained from 5 different scanning electron microscopy (SEM) images were respectively 1.5 µm and 20 nm. Then, the number of CNTs was calculated by matching the areas occupied by the CNTs on the substrate. The CNT areas were obtained from the SEM images by using ImageJ software, 4 as shown in Fig. S6. In order to have reliable data, at least 5 6
different SEM images taken from different regions were processed for each set of samples and the areas were averaged. The average areas covered by CNTs were 2.8%, 7.5%, 19.7%, 34.1%, and 78.5% for Sample VL, L, M, H, and VH, respectively. Figure S6. SEM images showing CNTs on Sample VL, L, M, H, VH whose CNT density is different by varying the spraying time of the CNT solution. The corresponding ImageJ data is shown underneath each SEM image. All scale bars in the SEM images indicate 2 μm. 7
CNTs bundles were created in a square of 25 25 µm 2 by randomly generating their orientations and positions with Fortran90 codes operated in the EOS system of Texas A&M supercomputing facility. An example of the generated CNTs is depicted in the inset of Fig. 4c. Here CNTs (indicated by lines) are depicted in a smaller square (5 5 µm 2 ) to clearly display the CNTs. The entire square was divided into multiple grids, as shown in Fig. S7. Here CNTs are indicated by the green lines and the grids are separated by the red lines passing through the tips of the CNTs. Note that only a portion (2.34 2.34 μm 2 ) of the entire square is depicted to clearly display the grids. Each grid consists of CNTs and PEDOT-Tos, and the thermal conductivity of each grid can be calculated by assuming a parallel connection of the PEDOT and CNTs, such as the parallel thermal resistors shown in Fig. S7. Thermal resistance (R) of the i-th grid can be expressed as: N 1 P N 1 C 1 R R R Grid, i j 1 PEDOT, j k 1 CNT, k where Np and Nc indicate the number of PEDOT and CNT segments. (4) Figure S7. Multiple grids were generated by intersecting the square. The grids were separated by horizontal red lines passing through the tips of the CNTs (green lines). The white background is PEDOT-Tos. Here a small portion of 2.34 2.34 μm 2 is displayed. Each grid can be considered as parallel connections of PEDOT-Tos and CNTs. The thermal resistance of each grid is serially connected, and the total thermal resistance can be obtained by summing all the thermal resistances. 8
This becomes equivalent to the volume weighted average of each component, which can be re-written with thermal conductivity (k) as: k k V k V (5) Grid, i CNT CNT, i PEDOT PEDOT, i where VCNT,i and VPEDOT,i represent the volume fractions of CNT and PEDOT in the i-th grid, respectively. The total thermal resistance of the square can be calculated as a serial connection of all grids. N R R (6) where N is the total number of the grids. The total thermal conductivity, k can be written as: 1 i 1 V Grid, i N Grid, i k i 1 kgrid, i where VGrid,i is the volume fraction of the i-th grid with respect to the total volume of the square. This series connection of the grids overestimates the thermal conductivity since this model assumes that CNTs are connected from the top to bottom without thermal resistance. For example, thermal conduction through CNTs that are physically disconnected at the interface of the grids is lower than that of connected CNTs, so this model considers the upper bound of thermal conductivity of the hybrid. In order to compensate this overestimation, the grids were also vertically divided, as shown in Fig. S8. Here we assumed that there is no heat flow across the vertical lines. Each vertical strip was independently calculated based on the skim for dividing the horizontal grids, as explained above. The width of the vertical strip was determined based on the measured electrical conductivity data. The electrical conductivities of PEDOT-Tos, CNT, and Sample L were measured to be 3.4 S/m, 2300 S/m and 13 S/m, respectively after 30-min reduction, and 9000 S/m, 43000 S/m, and 9200 S/m, respectively before the reduction. The matrix width that matches the electrical conductivity of Sample L with those of PEDOT-Tos and CNT only samples were found to be 390 nm before the reduction and 780 nm after reduction. Considering the assumption of the MC simulation and the uncertainty of the experiments, the averaged value (585 nm) was used for the MC simulation. Finally the thermal conductivity was calculated as a function of kcnt (effective thermal conductivity of CNT containing thermal contact resistance between CNTs 5 ) with kpedot=0.28 W/m-K (measured value in our work). We found that the thermal conductivity of Sample L (7) 9
matched the measurement results when kcnt is 60 W/m-K. Using kcnt of 60 W/m-K, the thermal conductivity of Samples VL and M was calculated to be 0.36 and 0.74 W/m-K, respectively, by changing the amount of CNT bundles accordingly. Based on the calculated thermal conductivity and measured electrical properties, ZT values of Sample VL and M were obtained, as shown in Fig. 4d. Figure S8. The grids were vertically divided by the blue lines. There is no heat transfer across the blue vertical lines, and the vertical strips were independently considered. The width of the strip was obtained from the fitting of electrical conductivity. References 1. NIST, Hall Effect Measurements, http://www.nist.gov/pml/div683/hall.cfm. 2. C. Liu, B. Lu, J. Yan, J. Xu, R. Yue, Z. Zhu, S. Zhou, X. Hu, Z. Zhang and P. Chen, Synth. Met., 2010, 160, 2481-2485. 3. V. Narayanunni, H. Gu and C. H. Yu, Acta Mater., 2011, 59, 4548-4555. 4. C. A. Schneider, W. S. Rasband and K. W. Eliceiri, Nat. Methods, 2012, 9, 671-675. 5. J. Hone, M. C. Llaguno, N. M. Nemes, A. T. Johnson, J. E. Fischer, D. A. Walters, M. J. Casavant, J. Schmidt and R. E. Smalley, Appl. Phys. Lett., 2000, 77, 666-668. 10