NANO LETTERS
Thermoelectric Behavior of Segregated-Network Polymer Nanocomposites
2008 Vol. 8, No. 12 4428-4432
Choongho Yu,* Yeon Seok Kim, Dasaroyong Kim, and Jaime C. Grunlan* Department of Mechanical Engineering, Texas A&M UniVersity, College Station, Texas 77843 Received August 1, 2008; Revised Manuscript Received September 21, 2008
ABSTRACT Segregated-network carbon nanotube (CNT)-polymer composites were prepared, and their thermoelectric properties were measured as a function of CNT concentration at room temperature. This study shows that electrical conductivity can be dramatically increased by creating a network of CNTs in the composite, while the thermal conductivity and thermopower remain relatively insensitive to the filler concentration. This behavior results from thermally disconnected, but electrically connected, junctions in the nanotube network, which makes it feasible to tune the properties in favor of a higher thermoelectric figure of merit. With a CNT concentration of 20 wt %, these composites exhibit an electrical conductivity of 4800 S/m, thermal conductivity of 0.34 W/m·K and a thermoelectric figure of merit (ZT) greater than 0.006 at room temperature. This study suggests that polymeric thermoelectrics are possible and provides the basis for further development of lightweight, low-cost, and nontoxic polymer composites for thermoelectric applications in the future.
Thermoelectric systems are very effective in harvesting electricity from waste heat or heat sources with small temperature gradients relative to environmental temperature.1,2 These heat sources are typically inadequate for power generation using conventional systems, but are often present in the environment or generated from the many power generating or consuming systems surrounding us in daily life. Furthermore, thermoelectric devices can be operated in a cooling mode and their simple leg-type structures, without moving parts, provide enormous advantages over conventional turbines, engines, and compressors, particularly for the applications that require robustness and silence. In addition, their high energy density (per unit weight and volume) is ideal for mobile power sources and cooling systems. For example, current state-of-the-art thermoelectric modules (without including heat sources) can have power densities more than one order higher than current typical diesel generator sets. This high power density and simple structure were enough to attract intense research efforts for improving their efficiency. However, until the recent advent of nanostructured Bi-Te based alloys,3-6 there has been almost no improvement in the efficiency of thermoelectric materials in the past 40 years.7 In this regard, it is important to explore new types of materials to further increase power density, as this would also have a significant impact on the performance of thermoelectrics (similar to efficiency en* The authors to whom correspondence should be addressed. E-mail:
[email protected] (C.Y.);
[email protected] (J.C.G.). 10.1021/nl802345s CCC: $40.75 Published on Web 10/30/2008
2008 American Chemical Society
hancement). For instance, if the Bi-Te alloys were replaced by similar performance polymeric materials, the power density of thermoelectric modules could be increased by approximately six to seven times as a result of the large difference in the densities of Bi2Te3 (∼7.86 g/cm3) and typical polymers (∼1.2 g/cm3). In addition, typical thermoelectric semiconductor materials are expensive and relatively difficult to process,8 impeding wide use of thermoelectric energy conversion. For instance, Te is very toxic and one of the rarest elements on earth. Polymer composites generally use simple manufacturing processes compared to semiconductor-based thermoelectric materials. This implies that polymer thermoelectrics could become competitive, even if their efficiency is less than that of current semiconductor materials. Polymers are intrinsically poor thermal conductors,9,10 which are ideal for thermoelectrics, but low electrical conductivity and thermopower have excluded them as feasible candidates for thermoelectrics in the past. However, recent progress in polymer and nanocomposite technology has made it possible to bring them into degeneratesemiconductor or metallic regimes by incorporating relatively small amounts of conductive fillers. The purpose of the present work is to demonstrate that polymer nanocomposites, particularly those made by a segregated network approach, can be viable as a lightweight and economical thermoelectric material. This method uses aqueous polymer emulsions, whose particles create excluded volume and essentially push
Figure 1. (a) Schematic of CNTs suspended in an aqueous emulsion. Gray spheres and red lines represent emulsion particles and CNTs, respectively. (b) Schematic of the emulsion-based composite after drying. The CNTs form a three-dimensional network along the surfaces of the spherical emulsion particles. SEMs of the cross-sections of 5, 5, 10, and 20 wt % CNT composites are respectively shown in panels c, d, e, and f after the composites were freeze-fractured (for instance, along the dotted line in panel b). The high-magnification SEM shown in panel d is a portion of the sample in c indicated by a yellow solid square. It clearly shows that CNTs (indicated by arrows) are wrapped around the emulsion particles (indicated by yellow dotted lines) rather than homogenously mixed. Denser CNTs were observed for higher concentration CNT composites. The scale bars in the SEMs indicate 1 µm.
nanoparticles (e.g., carbon nanotubes (CNTs)) into the interstitial space between them. This situation dramatically reduces the space available for the filler to form conductive networks, which results in a significant enhancement of electrical conduction with a relatively small amount of electrically conductive filler.11 Moreover, the manufacturing process is simple, scalable, and suitable for low-cost bulk synthesis. In this paper, an electrically insulating polymer matrix was chosen with CNTs as an electrically conductive filler in order to perform a feasibility study. CNT-polymer composite preparation, microstructure, and thermoelectric properties are described below. The composite matrix was made from a poly(vinyl acetate) (PVAc) homopolymer emulsion (Vinac XX210) (Air Products, Inc.) that was 55 wt % solids in water. A polymer emulsion is a stable suspension of solid, spherical polymer particles in water prior to film formation.12 The PVAc emulsion used here has a wide size distribution, 0.14-3.5 µm in diameter and an average diameter of ∼650 nm.10 XDNano Lett., Vol. 8, No. 12, 2008
grade CNTs (Carbon Nanotechnologies, Inc.), which are a mixture of metallic and semiconducting single-, double-, and triple-walled CNTs, were incorporated in the matrix. Gum arabic (GA) (Sigma-Aldrich, Co.) was used to stabilize the CNT in water.13 In order to prepare the composites, dry CNTs were combined with 2 wt % aqueous solutions of GA by sonication with a VirTis Virsonic 100 ultrasonic cell disrupter (SP industries, Inc.) for 15 min at 50 W. The CNT had a 1:1 weight ratio with GA. The PVAc emulsion and deionized water were then added to the GA/CNT mixture to produce an aqueous precomposite mixture containing 10 wt % total solids (except for the highest CNT concentration that was a 1.5 wt % solids mixture to reduce viscosity), as shown schematically in Figure 1a, followed by a 5-min sonication. For this study, eight different CNT concentrations, 0.5, 1, 2, 3, 4, 5, 10, and 20 wt %, in PVAc were prepared. These concentrations are based upon the total dry weight of the composite, which includes CNT, PVAc, and GA. In addition, 0 wt % CNT polymer films were also synthesized to compare 4429
thermal conductivities of the composites. Solid composites were made by drying aqueous mixtures in a 26 cm2 plastic mold for 2 days under ambient conditions and then for 24 h in a vacuum desiccator prior to testing in order to completely remove residual water. The thicknesses of the composites were measured to be 0.19, 0.20, 0.20, 0.20, 0.19, 0.18, 0.16, 0.24, and 0.078 mm for the 0, 0.5, 1, 2, 3, 4, 5, 10, and 20 wt % CNT composites, respectively. As shown in Figure 1b, the CNTs formed a three-dimensional network within the interstitial spaces between the emulsion particles. The samples were freeze-fractured for microstructural analysis (for instance, along the dotted line in Figure 1b). Figure 1c-f shows scanning electron micrographs (SEMs) of the cross section of 5, 5, 10, and 20 wt % CNT composites, respectively. A high-magnification image (Figure 1d) of a portion in Figure 1c (indicated by a yellow square) was inserted in order to highlight the presence of emulsion particles (indicated by dotted lines in Figure 1d) and chainlike CNT bundles (indicated by arrows in Figure 1d). It is clearly seen that CNTs wrapped around the emulsion particles in a network fashion rather than randomly distributed in the composite. Note that the typical length of an individual single wall CNT is ∼1 µm and the CNTs are often coated with a thin layer of GA.12 For simultaneous electrical conductivity and thermopower measurements, samples were cut into pieces (typically ∼30 mm × ∼7 mm) and suspended by using a thermal paste between two thermoelectric devices (∼20 mm apart) used for creating temperature gradients. Near both ends of the slender samples, four electrically conductive metal lines were made with silver paint for current-voltage (I-V) sweeping measurements. The I-V sweeping indicated the contact is ohmic. Electrical conductance was then obtained by taking the slope of the I-V curve and converted into electrical conductivity (σ) by multiplying a geometric factor. The thermoelectric devices, which are electrically connected in series to a power supply, have top and bottom sides that can be made to be cold and hot by passing an electrical current. By switching the top and bottom of one of the two devices and changing the supplying current, the temperature gradient across the sample can be altered. Two T-type thermocouples were attached to the metal lines on both ends of the sample with silver paint to simultaneously measure temperature differences and thermoelectric voltages (by using copper leads in the thermocouple) across the sample. It should be noted that the end of the sample is either cooled or heated, which cancels the thermoelectric voltage generated from the copper leads (the thermopower of copper at 300 K is also small, ∼1.83 µV/K.14). While the temperature gradient was varied up to ( 10 °C, thermoelectric voltages were measured so as to obtain thermopower (S) from the slope (S ) ∆VTE/∆T, where VTE and T are the thermoelectric voltage and temperature, respectively). Note that the electrical conductivity and thermopower measurements were performed with the same sample. As the two parameters are often strongly correlated, this minimizes the uncertainty that might be present due to the slight difference of microstructures and CNT concentra4430
Figure 2. Thermal conductivities of CNT-polymer composites at room temperature when the CNT concentrations are 0, 0.5, 1, 2, 3, 4, 5, 10, and 20 wt %, respectively.
tions over the whole sample. Another set of circular samples with 26.87 mm in diameter were prepared for thermal conductivity measurements. These samples were clamped between two cylindrical stainless steel rods whose ends were heated and cooled, respectively, for creating temperature gradients along the thickness direction of the sample. A series of five thermocouples were embedded along the axial direction of each rod and used to determine q and ∆T in the thermal conductance relation (Gth ) q/∆T, where Gth and q are thermal conductance per unit area (W/K·m2) and heat flux (W/m2), respectively). The temperature difference (∆T) across the sample was extrapolated from the series of temperature measurements and was typically ∼3 K. The thermal conductivity (k) of the sample was obtained from k ) t × Gth, where t is thickness of the sample. In order to prevent convection and radiation heat losses to the environment, the rods were wrapped by a thermal insulation cover. As shown in Figure 2, the thermal conductivity of the 0.5 and 1 wt % sample was slightly decreased compared to the 0 wt % sample, but it was more or less the same up to 5 wt % CNT. At 20 wt % CNT, a maximum thermal conductivity ∼1.5 times larger than that of the 0 wt % sample was observed. The decrease at 0.5 and 1 wt % is likely due to the phonon scattering by the CNTs.10 A tiny quantity of CNTs may act as impurities in the polymer matrix, but the relatively small change in thermal conductivity, even at high CNT concentration (10 and 20 wt %), is somewhat surprising based on the fact that the thermal conductivity of an individual CNT could be extremely high, up to several thousand W/m·K.15 The thermal conductivity (k) of a composite can be described from a parallel thermal resistor model as: k ) kmVm + kfVf
(1)
kf′
where km and are thermal conductivities of a matrix and a filler, and Vi represents the volume “fraction” of the indexed (i) material, m (matrix) and f (filler) (i.e., the volume of the indexed material divided by total volume). One can estimate that 20 vol % yields a composite thermal conductivity of ∼200 W/m·K as the thermal conductivity of the CNT and emulsion polymer can be respectively ∼103 W/m·K and ∼0.2 W/m·K. Note that the weight percent is close to the volume percent in this study (densities of PVAc and CNT are ∼1.2 Nano Lett., Vol. 8, No. 12, 2008
Figure 3. Electrical conductivity (indicated by red circles) and thermopower (indicated by blue squares) of 0.5, 1, 2, 3, 4, 5, 10, and 20 wt % CNT composites at room temperature. The red dashed line for the electrical conductivity is a three-parameter power fit (eq 2) over the CNT concentration from 0.5 to 20 wt %. The inset is a linear-log plot of the electrical conductivity as a function of the CNT wt %.
g/cm3 and ∼1.3 g/cm3, respectively).10 This large discrepancy in k likely comes from the many connections between CNTs along the long chain shown in Figure 1. They are often connected in series by weak van der Waals forces, which are expected to play a significant role in impeding phonon transport at the connected junctions. This may also be a rationale of many thermal conductivity measurement results of CNT bundles or mats that have yielded considerably lower values (down to a few W/m·K at room temperature) than theoretically predicted values.16-21 Therefore, k′f is smaller than the intrinsic property (kf) of the filler material (i.e., k′f < kf in the segregated network composite). This implies that the concentration of the electrically conductive filler can be increased without significantly raising the composite thermal conductivity, which is favorable for the enhancement of the thermoelectric performance that can be indicated by the thermoelectric figure of merit (ZT ≡ S2σ T/k). The benefit of the segregated network composite comes from the existence of the junctions that significantly impede thermal transport, but electronic properties can be maintained by a transport mechanism like hopping. Furthermore, the network structure allows most of the CNTs to contribute to the electrical conduction as opposed to homogeneous nanostructured composites, so as to maximize the influence from the filler. Figure 3 shows the electrical conductivity and thermopower of the 0.5, 1, 2, 3, 4, 5, 10, and 20 wt % CNT composites at room temperature. The electrical conductivity (σ) typically obeys a power law as a function of the conductive filler fraction:22 σ ) σ0|Vf - V*|p
(2)
where σ0 is a proportionality constant related to the intrinsic conductivity of the filler, and V* is the critical volume fraction Nano Lett., Vol. 8, No. 12, 2008
of the filler associated with the percolation threshold. Equation 2 fits the experimental data when σ0 ) 1.09 × 106 S/m, V* ) 0.0001, p ) 3.37, and R2 ) 0.9994 (coefficient of determination). In this case, fitting was done using wt % CNT instead of vol % because of density uncertainties for the CNT, but the results should be very similar because the densities of the polymer and CNT are close to one another. At 20 wt %, electrical conductivity is ∼4800 S/m, which is orders of magnitude higher than the typical 1-10 S/m values observed in more traditional nanotube-filled polymer composites with similar concentrations.23-26 The linear-log scale inset indicates that the electrical conductivity fits very well to the three-parameter power curve over the entire CNT loading curve (see inset of Figure 3). Nevertheless, the percolation threshold was not further tested with very low CNT concentration composites, as the focus of this study is on high conductivity samples. Despite this large increase in electrical conductivity, thermopower was more or less constant (40-50 µV/K). This value is close to the thermopower of a metallic CNT,15,27 as thermopower does not depend on filler dimensions and is strongly affected by more electrically conductive paths. This feature is ideal for tailoring thermoelectric properties because the electrical conductivity can be increased without significantly sacrificing thermopower, which is opposite of the behavior observed in bulk crystalline materials.28 From the values of the thermoelectric parameters shown in Figures 2 and 3, ZT of the composite with 20 wt % CNT was calculated to be ∼0.006 at 300 K, which is at least 6 times greater than those of the few previous studies with polymers,29,30 but smaller than those of typical semiconductor materials such as silicon (ZT ∼ 0.01 at 300 K).1,2,7,31 Nevertheless, ZT is expected to be significantly enhanced when the emulsion-based matrix is replaced by an intrinsically conductive polymer, and filler materials with excellent power factors (S2σ)29,32-35 are added to the composite by using the approach suggested in this study. For instance, the replacement of the electrically nonconducting matrix with poly(3,4-ethylenedioxythiophene) poly(styrenesulfonate) (PEDOT:PSS), whose electrical conductivity ranges typically between 102 and 104 S/m at room temperature depending on the type and concentration of dopants,36,37 would result in much higher electrical conductivity. PEDOT: PSS often exists as an aqueous suspension of polymer particles,37 much like a polymer emulsion. Furthermore, the polymer nanocomposite could be viable even with a low ZT, considering the high power density (lightweight), low cost, simple manufacturing process, and nontoxicity of polymeric materials. In summary, segregated-network CNT-polymer composites were prepared and their thermoelectric properties-thermal conductivity, electrical conductivity, and thermopower-were measured as a function of CNT concentration at room temperature. This study demonstrates that the electrical conductivity of a polymer composite can be dramatically increased by incorporating CNTs in a network fashion, while the thermal conductivity and thermopower were kept insensitive to the filler concentration. This behavior is believed to come from the thermally disconnected, but electrically connected, junctions that are present in the CNT networks that make it possible to tune the properties in favor of a higher thermoelectric figure of merit. Light-weight, low-cost, and nontoxic properties could make polymer composites 4431
viable, even if their efficiency is less than that of current semiconductor materials. Finally, it may be possible to achieve much greater improvement in the thermoelectric figure of merit for the polymer composite by using an intrinsically conductive polymer matrix and further varying filler materials and concentrations. Acknowledgment. The authors would like to acknowledge financial support from the Texas Engineering Experiment Station (TEES). J.C.G. would like to acknowledge additional support from the National Science Foundation under Award No. CMMI-0644055. References (1) Snyder, G. J.; Toberer, E. S. Nat. Mater. 2008, 7, 105. (2) Tritt, T. M.; Subramanian, M. A. MRS Bull. 2006, 31, 188. (3) Harman, T. C.; Taylor, P. J.; Walsh, M. P.; LaForge, B. E. Science 2002, 297, 2229. (4) Venkatasubramanian, R.; Siivola, E.; Colpitts, T.; O’Quinn, B. Nature 2001, 413, 597. (5) Hsu, K. F.; Loo, S.; Guo, F.; Chen, W.; Dyck, J. S.; Uher, C.; Hogan, T.; Polychroniadis, E. K.; Kanatzidis, M. G. Science 2004, 303, 818. (6) Poudel, B.; Hao, Q.; Ma, Y.; Lan, Y. C.; Minnich, A.; Yu, B.; Yan, X.; Wang, D. Z.; Muto, A.; Vashaee, D.; Chen, X. Y.; Liu, J. M.; Dresselhaus, M. S.; Chen, G.; Ren, Z. Science 2008, 320, 634. (7) Majumdar, A. Science 2004, 303, 777. (8) Winters, J. Mech. Eng. 2008, 130, 30. (9) Hu, M.; Yu, D.; Wei, J. Polym. Test. 2007, 26, 333. (10) Grunlan, J. C.; Kim, Y. S.; Ziaee, S.; Wei, X.; Abdel-Magid, B.; Tao, K. Macromol. Mater. Eng. 2006, 291, 1035. (11) Grunlan, J. C.; Mehrabi, A. R.; Bannon, M. V.; Bahr, J. L. AdV. Mater. 2004, 16, 150. (12) Keddie, J. L. Mater. Sci. Eng. R 1997, 21, 101. (13) Bandyopadhyaya, R.; Nativ-Roth, E.; Regev, O.; Yerushalmi-Rozen, R. Nano Lett. 2002, 2, 25. (14) Rowe, D. M., CRC Handbook of Thermoelectrics; CRC Press: Boca Raton, FL, 1995.
4432
(15) Yu, C.; Shi, L.; Yao, Z.; Li, D.; Majumdar, A. Nano Lett. 2005, 5, 1842. (16) Shi, L.; Li, D.; Yu, C.; Jang, W. Y.; Kim, D.; Yao, Z.; Kim, P.; Majumdar, A. J. Heat Transfer 2003, 125, 881. (17) Yu, C. Ph.D. Dissertation, University of Texas at Austin, 2004. (18) Hone, J.; Llaguno, M. C.; Nemes, N. M.; Johnson, A. T.; Fischer, J. E.; Walters, D. A.; Casavant, M. J.; Schmidt, J.; Smalley, R. E. Appl. Phys. Lett. 2000, 77, 666. (19) Hone, J.; Llaguno, M. C.; Biercuk, M. J.; Johnson, A. T.; Batlogg, B.; Benes, Z.; Fischer, J. E. Appl. Phys. A: Mater. Sci. Process. 2002, 74, 339. (20) Hone, J.; Whitney, M.; Piskoti, C.; Zettl, A. Phys. ReV. B 1999, 59, R2514. (21) Berber, S.; Kwon, Y. K.; Tomanek, D. Phys. ReV. Lett. 2000, 84, 4613. (22) Kirkpatrick, S. ReV. Mod. Phys. 1973, 45, 574. (23) Kymakis, E.; Amaratunga, G. A. J. J. Appl. Phys. 2006, 99. (24) Zhu, D.; Bin, Y.; Matsuo, M. J. Polym. Sci., Part B 2007, 45, 1037. (25) Grunlan, J. C.; Liu, L.; Kim, Y. S. Nano Lett 2006, 6, 911. (26) Ramasubramaniam, R.; Chen, J. Appl. Phys. Lett. 2003, 83, 2928. (27) Hone, J.; Ellwood, I.; Muno, M.; Mizel, A.; Cohen, M. L.; Zettl, A.; Rinzler, A. G.; Smalley, R. E. Phys. ReV. Lett. 1998, 80, 1042. (28) Yu, C.; Scullin, M. L.; Huijben, M.; Ramesh, R.; Majumdar, A. Appl. Phys. Lett. 2008, 92, 092118. (29) Hostler, S. R.; Kaul, P.; Day, K.; Qu, V.; Cullen, C.; Abramson, A. R.; X., Q.; Burda, C. Proc. Therm. Thermomech. Phenom. Electron. Syst. (ITHERM) 2006, 1400. (30) Levesque, I.; Gao, X.; Klug, D. D.; Tse, J. S.; Ratcliffe, C. I.; Leclerc, M. React. Funct. Polym. 2005, 65, 23. (31) Weber, L.; Gmelin, E. Appl. Phys. A: Mater. Sci. Process. 1991, 53, 136. (32) Shakouri, A.; Li, S. Proc. Int. Conf. Thermoelectr. 1999, 402. (33) Heremans, J. P.; Thrush, C. M.; Morelli, D. T.; Wu, M. C. Phys. ReV. Lett. 2002, 88. (34) Groenendaal, L. B.; Jonas, F.; Freitag, D.; Pielartzik, H.; Reynolds, J. R. AdV. Mater. 2000, 12, 481. (35) Kirchmeyer, S.; Reuter, K. J. Mater. Chem. 2005, 15, 2077. (36) Kim, J. Y.; Jung, J. H.; Lee, D. E.; Joo, J. Synth. Met. 2002, 126, 311. (37) Clevios, P. Formulation Guide; H.C. Starck Co.: Germany, 2008.
NL802345S
Nano Lett., Vol. 8, No. 12, 2008
