Increasing the thermoelectric power factor of polymer composites using a semiconducting stabilizer for carbon nanotubes

CARBON

5 0 ( 2 0 1 2 ) 8 8 5 –8 9 5

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/carbon

Increasing the thermoelectric power factor of polymer composites using a semiconducting stabilizer for carbon nanotubes Gregory P. Moriarty, Jamie N. Wheeler, Choongho Yu, Jaime C. Grunlan

*

Department of Mechanical Engineering, Texas A&M University, 3123-TAMU, College Station, TX 77843, USA

A R T I C L E I N F O

A B S T R A C T

Article history:

Poly(vinyl acetate) (PVAc) copolymer latex-based composites were prepared with

Received 23 June 2011

multi-walled carbon nanotubes (MWCNT), stabilized with sodium deoxycholate (DOC) or

Accepted 26 September 2011

meso-tetra(4-carboxyphenyl) porphine (TCPP). SEM images show that a segregated

Available online 2 October 2011

MWCNT network developed during drying, which resulted in relatively low percolation thresholds (1.62 and 2.17 wt.% MWCNT for DOC and TCPP, respectively). The electrical conductivity (r) of TCPP-stabilized composites is very similar to that of DOC-stabilized, while the thermopower (or Seebeck coefficient (S)) is five times as large. This enhanced thermopower suggests the MWCNT:TCPP/PVAc composite will have an order of magnitude greater power factor (S2r), which is an important measure of efficiency for thermoelectric materials (i.e., materials capable of converting a thermal gradient to a voltage). The thermal conductivity of these composites remains comparable to typical polymeric materials due to numerous tube–tube connections that act as phonon scattering centers. The universality of this approach was confirmed using much more electrically conductive double-walled carbon nanotube-filled composites that showed similar improvement with TCPP stabilization. It is possible that other porphyrin derivatives, or semiconducting molecules capable of stabilizing nanotubes in water, could be used to further enhance the Seebeck coefficient and improve the ability of these composites to convert waste heat into electricity. Ó 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

Thermoelectric devices are capable of harvesting electricity from environmental waste heat that generates a temperature gradient [1,2]. These temperature gradients are created by inefficient power sources (e.g., engines) or from systems consuming this power. At the atomic scale, this gradient causes electrons or holes in the material to diffuse from the hot side to the cold or vice versa. This movement of carriers creates an electric current that can be harnessed as useful voltage. Thermoelectric devices have the ability to increase the efficiency of any system without the need for moving parts and have numerous advantages over conventional systems (e.g.,

robustness and silence). The most useful traits of most thermoelectric systems are their high power density and relatively simple structure. Power density can be more than one order of magnitude higher for current thermoelectric systems than traditional diesel generator sets [3]. The thermoelectric figure of merit (ZT) is a common measure of a material’s energy conversion efficiency: ZT ¼

S2 rT k

ð1Þ

where S (in V/K) is the thermopower (or Seebeck coefficient), r (in S/m) is the electrical conductivity, k (in W/m K) is the thermal conductivity, and T (in K) is the absolute measurement

* Corresponding author: Fax: +1 979 845 3027. E-mail address: [email protected] (J.C. Grunlan). 0008-6223/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.carbon.2011.09.050

886

CARBON

5 0 ( 2 0 1 2 ) 8 8 5 –8 9 5

temperature [4]. Because thermal conductivity tends to be the least variable of the properties that contribute to ZT, the power factor (S2r) is commonly reported in its place as a simpler measure of thermoelectric efficiency [2]. Semiconductors, in general, have the best combination of properties to achieve a good thermoelectric efficiency. The most efficient thermoelectric material is bismuth telluride, which has a power density of up to 0.2 kW/kg, but is expensive due to the scarcity of tellurium [5–9]. Bi2Te3 and its derivatives have been shown to achieve ZT values around 1 [1,10], corresponding to a Carnot efficiency of approximately 8% [11]. Unfortunately, these semiconductor alloys contain rare and expensive elements, are difficult to process and have not significantly improved ZT in 40 years [12]. More recently, polymer-based materials have been studied as lower cost, more environmentally friendly alternatives to semiconductor thermoelectrics [3,13–17]. Segregated networks of carbon nanotubes in a poly(vinyl acetate), latex-based matrix have calculated ZT values as high as 0.02 [13], which is at least an order of magnitude greater than any other fully organic systems [11,14–23]. The polymer matrix has a low thermal conductivity (<0.4 W/m K), with desirable mechanical properties and low density, that is nearly an order of magnitude lower than Bi2Te3. Using a polymer emulsion (also known as a latex) as the composite matrix allows highly conductive composites to be prepared from water-based suspensions [24]. The emulsion exists as an aqueous suspension of solid polymer particles with a diameter of 0.1–1 lm [25]. Carbon nanotubes (CNT), which have high electrical conductivity, but relatively low Seebeck coefficient [26], are forced into the interstitial positions between these polymer particles. This excluded volume effect creates an electrically conductive network with a relatively small amount of filler [24]. The electrical conductivity of these polymer composites generally obeys a power law [27]: r ¼ r0 ðV  Vc Þs

ð2Þ

where r0 is a proportionality constant related to the effective intrinsic conductivity of the CNT in the composite, V is the volume fraction of CNT, Vc is the critical volume fraction of CNT to produce the conductive network (i.e., the percolation threshold), and s is the power law exponent [28]. In typical bulk semiconductors, increasing the electrical conductivity often results in a decrease in the Seebeck coefficient [1]. In contrast, it is possible to simultaneously increase the electrical conductivity and keep the Seebeck coefficient and thermal conductivity relatively constant with a polymer composite [3]. This behavior can best be illustrated as creating electrically conductive junctions that are thermally disconnected, resulting in energy transfer barriers. In order to make water-based composites containing carbon nanotubes, stable suspensions must first be prepared. Carbon nanotubes require stabilizing agents because of their hydrophobic and highly entangled nature that prevents complete dispersion and/or exfoliation in water [29,30]. There are several different types of stabilizers, each having a unique influence on the electron transport across junctions, that have been used to disperse CNT in water. Surfactants [31–34], polymers [35–38], and inorganic nanoparticles [39,40] have all been successfully used for this purpose. Intrinsically conductive

polymer stabilizers, in particular, can dramatically increase composite electrical conductivity [13]. Increasing conductivity with these stabilizers, and simultaneously increasing thermopower, will lead to improved polymer-based thermoelectric materials. In the present work, a semiconducting molecule, mesotetra(4-carboxyphenyl) porphine (TCPP), was used to stabilize multi-walled carbon nanotubes (MWCNT) and double-walled carbon nanotubes to produce segregated network composites, in an effort to increase thermopower. It has been shown that other porphyrin derivatives interact and stabilize carbon nanotubes due to their conjugated chemistry that promotes p–p overlap [41]. The capability of porphyrin derivatives to act as a surfactant, and also enhance electrical conductivity (or Seebeck coefficient) [42], suggests they can be used to make composites with improved ZT. A composite made with poly(vinyl acetate) latex and 12 wt.% MWCNT exhibits an increase in Seebeck coefficient, from 8 to 28 lV/K, when the stabilizer is changed from sodium deoxycholate (DOC) to TCPP. The porphyrin stabilizer is not as effective as DOC, so there is a slight reduction in electrical conductivity at a given MWCNT concentration. The two stabilizing agents do not interact with the latex, as evidenced by an unaltered glass transition temperature; however, the porphyrin stabilized composite reaches a critical pigment volume concentration well before the DOC-stabilized system. The universality of this approach to increase Seebeck coefficient is demonstrated by replacing MWCNT with double-walled carbon nanotubes (DWCNT) in the composites. S increases from 71 to 78 lV/K when the stabilizer is changed from DOC to TCPP for DWCNT-filled/PVAc composites. Composite thermal conductivity did not significantly change from that of unfilled polymeric materials (0.2–0.4 W/m K) as the stabilizer and CNT were exchanged. This use of a semiconductor stabilizer provides a new tool for enhancing the thermoelectric properties of polymer nanocomposites and provides greater potential for their use in harvesting waste heat, especially from places where semiconductors would be impractical (e.g., painted surfaces).

2.

Experimental

2.1.

Materials

A vinyl acetate/acrylic copolymer (PVAc) emulsion (Rovace TM 86 supplied by Rohm and Haas, Spring House, PA), that is 54.7 wt.% solids in water, was used as the composite matrix starting material. This PVAc emulsion has an average particle diameter of 346 nm ± 7 nm. Multi-walled carbon nanotubes (Baytubes C 150P provided by Bayer MaterialScience, Leverkusen, Germany), with an average diameter of 14 nm and a length of 1–10 lm, were used as the model electrically conductive filler. Double-walled carbon nanotubes (DWCNT) (XBC1001 purchased from Continental Carbon Nanotechnologies, Inc., Houston, TX) were used as higher conductivity filler. Sodium deoxycholate (DOC) (Sigma–Aldrich, Saint Louis, MO) or meso-tetra (4-carboxyphenyl) porphine (TCPP) (Frontier Scientific, Logan, UT) were used to stabilize the nanotubes in water during composite preparation.

CARBON

2.2.

5 0 (2 0 1 2) 8 8 5–89 5

Composite preparation

MWCNT:stabilizer weight ratios of 1:1.57 and 1:3 were prepared using aqueous solutions of 3 wt.% DOC and 2.67 wt.% TCPP, respectively. These weight ratios were chosen to maintain equal moles of surfactant per gram of MWCNT. These aqueous suspensions were then sonicated with a VirTis Virsonic 100 ultrasonic cell disrupter (SP Industries Inc., Warminster, PA) for 10 min at 50 W in an ice water bath. The PVAc emulsion and deionized water were adjusted (using a 0.1 M NaOH solution) to pH 10, as this was the pH value of MWCNT:DOC solutions (and of the TCPP molecule needed in order to be stable in water). The polymer emulsion and pHadjusted deionized water were then added to the MWCNT:stabilizer mixture and sonicated again for another 10 min at 50 W. This final aqueous suspension contained 5 wt.% total solids. Composites with seven different MWCNT concentrations (2.5, 3, 4, 5, 7, 10, and 12 wt.%) were prepared by drying suspensions in a 26 cm2 plastic mold for 2 days under ambient conditions and then for 24 h in a vacuum desiccator. This was done to ensure that all residual moisture was removed prior to testing. Concentrations are based upon the dry weight of PVAc, MWCNT, and surfactant solids used in the composite. Composites containing 7, 10, and 12 wt.% DWCNT were prepared the same way, but with aqueous suspensions containing 2.5 wt.% total solid to reduce viscosity.

2.3.

Composite characterization

The sheet resistance of these composites was measured with a four-point probe (Signatone S-301 series, Gilroy, CA), which was then converted to electrical conductivity (i.e., the product of inverse sheet resistance and film thickness). Five measurements were taken on each side (top and bottom surfaces of the composites) to confirm that a given specimen was isotropic. To measure the Seebeck coefficient, samples were cut into a rectangular shape (20 mm in length and 3 mm in width) and measured with a home-built, shielded four-point probe apparatus, equipped with a Keithley 2000 Multimeter (Cleveland, OH) and a GW PPS-3635 power supply (Good Will Instrument Co., Ltd) and operated with a Labview (National Instruments, Austin, TX) interface [3]. Glass transition temperatures and storage moduli were measured with a Q800 Dynamic Mechanical Analyzer (DMA) from TA Instruments (New Castle, DE). The composites were cut into strips (27 mm in length and 4 mm in width) and measured in tensile mode, with an amplitude of oscillation of 15 lm. Temperature was ramped at a rate of 5 °C/min, from 70 to 70 °C, during testing. The electron micrographs of composite cross-sections were taken with an FEI Quanta 600 FE-SEM (Hillsboro, OR). Samples were soaked in liquid nitrogen and freeze fractured by hand, then sputter coated with 5 nm of platinum prior to imaging. During imaging, the accelerating voltage was 10 kV, with a spot size of 3.0 nm and a working distance of approximately 10 nm. Thermal conductivity was measured in the film thickness direction with a homemade ASTM D5470 standard setup. This setup was constructed with two brass blocks connected to an electrical heater and circulating water cooling bath to maintain a constant temperature gradient. The electrical heater was set to a constant output of 50 °C, while

887

the circulating water was set at 0 °C. Meter bars, comprised of one inch diameter stainless steel, were then attached to the brass blocks in order to extrapolate a linear array of thermocouples to the exact temperatures on both sides of the sample being measured. A thermal paste was used to help increase the thermal conductivity of the thermal interface between the meter bars and sample. This system has an error of 18%, as noted by the ASTM standard.

3.

Results and discussion

3.1.

Composite microstructure

Fig. 1(a) and (b) schematically shows the dispersing behavior of DOC and TCPP, respectively. Both stabilizers exfoliate the MWCNT in solution by adsorbing to their surfaces and changing them from hydrophobic to hydrophilic. DOC was chosen as a reference stabilizer because it is already known to effectively exfoliate CNT in water [43–45]. Porphyrin complexes have also been used as stabilizing agents for CNT. These complexes are believed to attach to nanotube surfaces through electrostatic [46] and p–p interactions [47], and also via axial coordination [48]. The chemical structures of each stabilizer reveal that DOC is an insulator and TCPP, with its conjugated backbone, is capable of conducting electrons. Fig. 1(c) and (d) schematically shows the formation of a segregated network upon drying of a polymer emulsion with the addition of a MWCNT:stabilizer suspension. The polymer particles exclude volume that the MWCNT could otherwise occupy, thereby forcing the nanotubes into the interstitial spaces between them. As shown in Fig. 1, this excluded volume ultimately results in the formation of a segregated network when V > Vc. Fig. 2 shows SEM cross-sectional micrographs of the 1:1.57 MWCNT:DOC/PVAc composites with varying MWCNT concentrations. The 2.5 wt.% MWCNT composite (Fig. 2(a)) does not clearly show the expected segregated network, which may be due to the sample being too close to the percolation threshold. There are relatively few long nanotube pathways in a composite with V near Vc, making them difficult to find with SEM. At higher magnification (Fig. 2(b)), the MWCNT are shown to be aggregating in bundles by the polymer particles instead of being more evenly dispersed throughout the available interstitial space. Even with stabilizer present, strong interactions exist among the nanotubes. Increasing the ratio of stabilizer to nanotube would likely improve dispersion, but it would be expected to simultaneously diminish electrical conductivity due to blocking of direct tube-to-tube contacts. When the MWCNT concentration is increased to 10 wt.% (Fig. 2(c)), the segregated network becomes more visible. The polymer forms relatively large domains (>10 lm) of many coalesced particles that the MWCNT forms a network around. At higher magnification (Fig. 2(d)), designated by the dotted-line box in Fig. 2(c), the MWCNT are shown to be located in the regions between the large polymer particle domains. The MWCNT do not appear to have significant interaction with the polymer, which contributes to higher conductivity. It is important to note that as the MWCNT concentration is increased, the porosity of the network also in-

888

CARBON

5 0 ( 2 0 1 2 ) 8 8 5 –8 9 5

Fig. 1 – The dispersed MWCNT, when mixed with polymer emulsion particles, forms a three-dimensional network in the interstitial positions between the polymer particles (a) and (b) show representations of the dispersed MWCNT in the two stabilizing agents, DOC and TCPP, (c) and (d) illustrate the formation of a segregated network upon drying of the water-based polymer emulsions.

creases. These microvoids form when the polymer particles can no longer fill the gaps between MWCNT and contribute to degradation of mechanical properties at high concentration. When the filler is changed to DWCNT, at a concentration of 10 wt.% (Fig. 2(e)), the network appears more uniform and detached from the matrix. A possible reason for this could be due to the smaller DWCNT not being stabilized as well by the DOC, allowing more bundles to form. There are still many tube-tube junctions (Fig. 2(f)) that allow for increased electrical conductivity. Fig. 3 shows SEM cross-sectional images of the MWCNT:TCPP/PVAc system at varying MWCNT concentrations. Fig. 3(a) is a 2.5 wt.% MWCNT composite, with Fig. 3(b) as the magnified image of the dotted-line box in Fig. 3(a). This system looks very similar to the MWCNT:DOC/ PVAc in Fig. 2. The 2.5 wt.% MWCNT composite is again near the percolation threshold, so there is little evidence of a segregated network. The same aggregation behavior is also observed, but the MWCNT looks to be more evenly dispersed. Fig. 3(c) shows the 10 wt.% MWCNT composite, which shows the true segregated network. As seen in the MWCNT:DOC/ PVAc system (Fig. 2(d)), porosity amongst the MWCNT can be observed at higher magnification (Fig. 3(d)). The pores seem more extensive here, which is believed to be related to the poorer stabilization by the higher molecular weight TCPP

stabilizer. As already mentioned, this porosity becomes a factor in the mechanical properties that will be discussed in the next section and is linked to the critical pigment volume concentration (CPVC) phenomenon [24,49–52]. The microscopic voids seem to have begun connecting with each other, leading to large scale defects. This porosity is a significant issue for segregated networks due to the inability of the polymer to effectively fill voids between particles. Beyond the CPVC, often below 10 vol.% for segregated networks, porosity becomes extensive and begins to degrade composite mechanical behavior [24,49,53]. This porosity is also seen when DWCNT is used as the electrically conductive filler (Fig. 3(e)). The DWCNT appears to be better stabilized by the TCPP, when compared to the DOC-stabilized system. A reason for this behavior could be due to the larger TCPP molecule being able to sterically stabilize the DWCNT.

3.2.

Thermo-mechanical behavior

Fig. 4 shows glass transition temperatures (Tg) for the MWCNT:DOC/PVAc and MWCNT:TCPP/PVAc composites. These values were taken as the inflection point of the loss modulus curves measured with DMA. The Tg of neat PVAc copolymer (0 wt.% MWCNT), is approximately 10 °C. As the MWCNT concentration increases (Fig. 4(a)), the Tg increases

CARBON

5 0 (2 0 1 2) 8 8 5–89 5

889

Fig. 2 – SEM cross-sectional images of 1:1.57 CNT:DOC composites containing 2.5 wt.% MWCNT (a) 10 wt.% MWCNT (c) and 10 wt.% DWCNT (e), (b), (d), and (f) are greater magnification images of (a), (c), and (e), respectively.

only slightly. This confirms that the MWCNT network does not interact strongly with the PVAc matrix. If there was a strong interaction between the filler and matrix, the Tg would be expected to increase more dramatically with increasing filler concentration due to the restriction of polymer chains [54–59]. In the MWCNT:TCPP/PVAc system (Fig. 4(b)), the Tg increases to 20 °C at 7 wt.% MWCNT, with a drop to 15 °C at 10 wt.% MWCNT. This increase may be attributed to some sort of molecular interaction of the TCPP with PVAc. Large error bars on these Tg values, especially at higher MWCNT content, are likely due to the porosity in the structure. Even with TCPP, the Tg increase is very modest and again suggests that the MWCNT has only weak interaction with the PVAc matrix. The storage moduli, measured at 60 °C, are shown in Fig. 5 for the DOC and TCPP stabilized systems. As expected, modulus increases with increasing MWCNT concentration

in the MWCNT:DOC/PVAc system (Fig. 5(a)), from approximately 6–12 GPa (with 10 wt.% MWCNT). The storage modulus peaks at 14 GPa with 7 wt.% MWCNT, then decreases at 10 wt.% (Fig. 5(b)). A more classical CPVC phenomenon is observed in the MWCNT:TCPP/PVAc system. This behavior was also seen in the Tg results (Fig. 4(b)). Pores are very apparent in the composite containing 10 wt.% MWCNT (Fig. 3(c)), which act as zero modulus filler. A CPVC is not observed in the MWCNT:DOC/PVAc system because the polymer matrix can still envelop the filler/stabilizer mixture. Better stabilization of MWCNT by DOC, relative to TCPP, is believed to be the primary reason for this difference. As MWCNT is better dispersed in the matrix, porosity will require a greater filler concentration to develop (i.e., higher CPVC). Latex-based films filled with 18 or 36 wt.% TCPP, in the absence of MWCNT, exhibited the same storage modulus as unfilled latex (7 GPa).

890

CARBON

5 0 ( 2 0 1 2 ) 8 8 5 –8 9 5

Fig. 3 – SEM cross-sectional images of 1:3 CNT:TCPP/PVAc composites containing 2.5 wt.% MWCNT (a), 10 wt.% MWCNT (c), and 10 wt.% DWCNT (e). The highlighted region, marked by a dotted box in (a), is enlarged in (b). Higher magnification images of (c) and (e) are shown in (d) and (f), respectively.

Fig. 4 – Glass transition temperatures, as a function of MWCNT concentration, for DOC (a) and TCPP (b) stabilized systems.

CARBON

5 0 (2 0 1 2) 8 8 5–89 5

891

Fig. 5 – Storage moduli measured at 60 °C, as a function of MWCNT concentration, for DOC (a) and TCPP (b) stabilized systems.

3.3.

Transport properties

Fig. 6 shows the thermal conductivities of the 7, 10, and 12 wt.% CNT:DOC/PVAc and CNT:TCPP/PVAc composites. Unlike electrical conductivity, the thermal conductivity has been shown to be relatively insensitive to CNT and stabilizer concentration [13]. Thermal conductivities range from approximately 0.18 W/m K, for the MWCNT:DOC/PVAc system, to 0.22 W/m K, for the MWCNT:TCPP/PVAc system. When 12 wt.% MWCNT is replaced by DWCNT, the thermal conductivity of the DOC composite is nearly identical to the TCPP composite at 0.17 W/m K. The thermal conductivity (k) of a composite can best be described by a parallel resistor model: k ¼ km Vm þ kf Vf

ð3Þ

where V is the volume fraction and subscripts, m and f, stand for the matrix and filler, respectively. It is important to note that weight fractions were used in place of volume fractions due to the uncertain density of MWCNT (and other composite ingredients) [60]. Eq. (3) provides a maximum thermal conductivity by maximizing the contribution of the conductive filler (assuming perfect contact between each CNT). This assumption is not entirely correct, as a perfect contact between each tube is impossible for these composites. Thermal

Fig. 6 – Thermal conductivity of DOC and TCPP stabilized systems containing 7, 10, and 12 wt.% MWCNT and DWCNT.

conductivity of the matrix (km) (0.2 (W/m K)) is much lower than that of a single carbon nanotube (kf) (1000 (W/m K)) [3,26,61,62], which would be expected to result in a large increase in composite thermal conductivity as the CNT volume concentration is increased. At 12 wt.% CNT, the predicted thermal conductivity of the composite would be 120 W/ m K, which is nearly three orders of magnitude greater than experimental results (0.17–0.22 W/m K). There is a dramatic difference between the theoretical value of a single MWCNT and the experimental value of the bulk measurement. This discrepancy is due to the numerous high thermal contact resistances between the tubes themselves, resulting in a reduced bulk measurement (15(W/m K)) [63,64]. If this value was chosen instead of the theoretical value, the thermal conductivity would be roughly 1.76 W/m K, which is still an order of magnitude different than those reported experimentally [13,65,66]. The stabilization of CNT by DOC and TCPP also suppress the thermal conductivity by blocking tube-to-tube junctions, creating less favorable pathways for phonon transport. These organic stabilizers can also act as phonon scattering centers because they are embedded in the composite alongside the CNT. This can be further explained by the increased thermal resistance, caused by the poor phonon coupling in vibrational modes of the polymer-filler and filler-filler at the interface, called the Kapitza resistance [67–71]. Other possible reasons for such a low thermal conductivity include gaps present between adjacent tubes due to misalignment, differing tube diameters and lengths, morphology of the CNTs, and defects introduced to the CNTs by functionalizing them during sonication with the dispersant [72–74]. Fig. 7 shows electrical conductivity as a function of MWCNT concentration for the DOC and TCPP-stabilized composites. Electrical conductivity increases exponentially at lower MWCNT concentrations for both systems. A solid line fit to this data was performed using the classical percolation power law (Eq. (2)). Weight fractions are used here instead of volume fractions as already explained above. The percolation threshold for the MWCNT:DOC/PVAc system was calculated to be 1.6 wt.% MWCNT, while the MWCNT:TCPP/PVAc threshold was found to be approximately 2.2 wt.% MWCNT.

892

CARBON

5 0 ( 2 0 1 2 ) 8 8 5 –8 9 5

These results support the earlier assessment of composite microstructure at low MWCNT concentrations. Higher percolation threshold suggests higher CPVC due to a greater MWCNT concentration required for strong network formation. Below 10 wt.% MWCNT, the DOC-stabilized composites exhibit slightly greater electrical conductivity due to better nanotube stabilization/exfoliation. The highest electrical conductivity of the MWCNT:TCPP/PVAc system was found to be approximately 128 S/m, at 12 wt.% MWCNT, which is 88% greater than the MWCNT:DOC/PVAc system. The increased electrical conductivity at lower MWCNT concentrations, with DOC stabilization, can be attributed to the tighter junctions that are formed relative to the bulkier TCPP [75]. Control films containing 18 or 36 wt.% TCPP, without any MWCNT, were too insulating (r < 0.01 S/m) to be measured using the standard four point probe apparatus. Although electrical conductivity for a given MWCNT concentration is quite similar between the two stabilizers, the TCPP stabilized composites have a much larger Seebeck coefficient (Fig. 8(a)). In both MWCNT:DOC/PVAc and MWCNT:TCPP/PVAc composites, only the 7, 10, and 12 wt.% MWCNT samples were evaluated due to the need for relatively high conductivity for measurement. These composites exhibit an R2 fit of 0.98 or higher for the voltage–temperature plot (not shown), which suggests an accurate Seebeck measurement. Seebeck coefficients range from approximately

8 lV/K, for the MWCNT:DOC/PVAc system, to 28 lV/K, for the MWCNT:TCPP/PVAc system. These values, which are much lower than those reported for composites filled with single-walled carbon nanotubes (SWCNT) (40–60 lV/K) [3], follow typical semiconductor behavior, where a higher electrical conductivity is accompanied by a lower thermopower and vice versa. The increased S with TCPP stabilization can be attributed to either a lower carrier concentration or larger effective mass of the carriers relative to DOC [1]. When 12 wt.% MWCNT is replaced by DWCNT, the electrical conductivity of the DOC composite is 1474 S/m and 7108 S/m for the TCPP composite. DOC does not provide as great an increase as TCPP because the DWCNT is better stabilized, causing the insulating nature of the stabilizer to inhibit the electron transfer from tube-to-tube. DWCNT, much like SWCNT, has a much higher S than MWCNT. For 12 wt.% DWCNT composites, the Seebeck coefficient is 71 lV/K with DOC. This value increases to 78 lV/K when TCPP replaces DOC as the stabilizer. The increased Seebeck coefficient with TCPP would increase ZT by nearly 20%, without even considering the simultaneous increase in conductivity for the DWCNT composite. DWCNT exhibits these increased properties due to the fact that either one (or both) of the tubes are metallic in character, while the other is semiconducting, which creates a more highly conductive material. The increased electrical conductivity could also be explained by the DWCNTs having

Fig. 7 – Electrical conductivity as a function of MWCNT concentration for the DOC (a) and TCPP (b) systems. The solid line in both graphs is the percolation power law fit, with the percolation thresholds provided as insets.

Fig. 8 – Seebeck coefficients (a) and power factors (b) of DOC and TCPP stabilized systems containing 7, 10, and 12 wt.% MWCNT and DWCNT.

CARBON

5 0 (2 0 1 2) 8 8 5–89 5

a higher structural stability, creating highly conductive p-conjugated pathways that remain undisturbed on the inner tube during functionalization [76–79]. The increase in Seebeck coefficient with the addition of this semiconducting stabilizer results in an increase in the power factor (S2r) (Fig. 8(b)). The power factor of these MWCNT-filled composites increases more than an order of magnitude, from 0.004 to 0.1 lW/m K2, when TCPP is replaced by DOC. A more modest increase is observed for the DWCNTfilled composites, with S2r rising from 7.53 to 42.8 lW/m K2. This example demonstrates the significant improvement in thermoelectric performance that occurs with seemingly small changes in S and r. Higher power factors have already been observed with higher nanotube concentrations, using intrinsically conductive PEDOT:PSS as the stabilizer [13]. Much like with bulk semiconductors, a compromise between electrical conductivity and the Seebeck coefficient will produce the highest ZT values. Table S1 summarizes all of the transport properties of these composites.

4.

Conclusion

Two series of latex-based segregated-network polymer composites were prepared with insulating (DOC) or semiconducting (TCPP) stabilizing agents for carbon nanotubes. Composite microstructure, mechanical, and transport properties were evaluated in an effort to understand the influence of using a porphyrin as the stabilizer. The glass transition temperatures for both systems were relatively unaltered with changing MWCNT concentration, which suggests that there was little interaction between the polymer matrix and the stabilized nanotubes. Thermal conductivities were relatively unaffected by stabilizer, or CNT concentration, and they were below those of typical polymeric materials (0.2–0.4 W/m K). MWCNT:DOC/PVAc composites exhibited a lower percolation threshold (1.6 wt.% MWCNT) than MWCNT:TCPP/PVAc (2.2 wt.% MWCNT). This threshold difference is attributed to the latex particles being able to envelop the better stabilized MWCNT:DOC/PVAc more effectively than the more weakly stabilized MWCNT:TCPP/PVAc. Seebeck coefficients were relatively unaffected by MWCNT concentration, but TCPP-stabilized composites had S values that were five times as large as DOC-stabilized. S and r both increased for composites containing DWCNT instead of MWCNT, further demonstrating the utility of semiconductor stabilizers for improving thermoelectric behavior. These improvements can be explained by an increase in the effective mass of the charge carriers in the TCPP molecule when compared to the DOC. Further improvements in Seebeck coefficient and/or electrical conductivity are expected with higher DWCNT concentration and the use of an intrinsically conductive polymer stabilizer (e.g., PEDOT:PSS) in combination with TCPP (or other semiconducting stabilizers).

Acknowledgements The FE-SEM acquisition was supported by the NSF grant DBI0116835, the VP for Research Office, and the TX Eng. Exp. Station. The authors thank Dr. James Batteas and Albert Wan for

893

their insightful discussions concerning TCPP and Dr. Diane Vaessen for help identifying an appropriate latex for this study. The authors thank the US Air Force Office of Scientific Research (Grant No. FA9550-09-1-0609), under the auspices of Dr. Charles Lee, for financial support of this work.

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.carbon.2011.09.050.

R E F E R E N C E S

[1] Snyder GJ, Toberer ES. Complex thermoelectric materials. Nat Mater 2008;7(2):105–14. [2] Tritt TM, Subramanian MA. Thermoelectric materials, phenomena, and applications: a bird’s eye view. MRS Bull 2006;31(3):188–94. [3] Yu C, Kim YS, Kim D, Grunlan JC. Thermoelectric behavior of segregated-network polymer nanocomposites. Nano Lett 2008;8(12):4428–32. [4] Rowe DM. Handbook of thermoelectrics. Boca Raton: CRC Press; 1995. [5] Harman TC, Taylor PJ, Walsh MP, LaForge BE. Quantum dot superlattice thermoelectric materials and devices. Science 2002;297(5590):2229–32. [6] Venkatasubramanian R, Siivola E, Colpitts T, O’Quinn B. Thinfilm thermoelectric devices with high room-temperature figures of merit. Nature 2001;413(6856):597–602. [7] Hsu KF, Loo S, Guo F, Chen W, Dyck JS, Uher C, et al. Cubic AgPbmSbTe2+m: bulk thermoelectric materials with high figure of merit. Science 2004;303(5659):818–21. [8] Poudel B, Hao Q, Ma Y, Lan YC, Minnich A, Yu B, et al. Highthermoelectric performance of nanostructured bismuth antimony telluride bulk alloys. Science 2008;320(5876):634–8. [9] Winters J. Heat to power, directly. Mech Eng 2008;130(8):30–3. [10] Majumdar A. Thermoelectricity in semiconductor nanostructures. Science 2004;303(5659):777–8. [11] Goldsmid GJ. In: 17th int conf on thermoelectrics. 1998. p. 25– 8. [12] Rosi FD. Thermoelectricity and thermoelectric power generation. Solid State Electron 1968;11(9):833–68. [13] Kim D, Kim Y, Choi K, Grunlan JC, Yu CH. Improved thermoelectric behavior of nanotube-filled polymer composites with poly(3,4-ethylenedioxythiophene) poly(styrenesulfonate). ACS Nano 2010;4(1):513–23. [14] Yoshino K, Morita S, Yin XH, Onoda M, Yamamoto H, Watanuki T, et al. Electrical property of polypyrroleinsulating polymer composite. Synth Met 1993;57(1):3562–5. [15] Jousseaume V, Morsli M, Bonnet A, Tesson O, Lefrant S. Electrical properties of polyaniline–polystyrene blends above the percolation threshold. J Appl Polym Sci 1998;67(7):1205–8. [16] Anilkumar KR, Parveen A, Badiger GR, Prasad MVNA. Thermoelectric power factor for polyaniline/molybdenum trioxide composites. Ferroelectrics 2009;386:88–93. [17] Levesque I, Gao X, Klug DD, Tse JS, Ratcliffe CI, Leclerc M. Highly soluble poly(2,7-carbazolenevinylene) for thermoelectrical applications: from theory to experiment. React Funct Polym 2005;65(1–2):23–36. [18] Yao Q, Chen LD, Zhang WQ, Liufu SC, Chen XH. Enhanced thermoelectric performance of single-walled carbon nanotubes/polyaniline hybrid nanocomposites. Acs Nano 2010;4(4):2445–51.

894

CARBON

5 0 ( 2 0 1 2 ) 8 8 5 –8 9 5

[19] Abramson AR, Kim WC, Huxtable ST, Yan HQ, Wu YY, Majumdar A, et al. Fabrication and characterization of a nanowire/polymer-based nanocomposite for a prototype thermoelectric device. J Microelectromech S 2004;13(3):505–13. [20] Kymakis E, Amaratunga GAJ. Electrical properties of singlewall carbon nanotube-polymer composite films. J Appl Phys 2006;99(8):84302. [21] Das NC, Liu YY, Yang KK, Peng WQ, Maiti S, Wang H. Singlewalled carbon nanotube/poly(methyl methacrylate) composites for electromagnetic interference shielding. Polym Eng Sci 2009;49(8):1627–34. [22] Chang KC, Jeng MS, Yang CC, Chou YW, Wu SK, Thomas MA, et al. The thermoelectric performance of poly(3,4-ethylenedi oxythiophene)/poly(4-styrenesulfonate) thin films. J Electron Mater 2009;38(7):1182–8. [23] Hostler SR, Kaul P, Day K, Qu V, Cullen C, Abramson AR. In: ITHERM ‘06 10th intersoc conf 2006. p. 1400. [24] Grunlan JC, Mehrabi AR, Bannon MV, Bahr JL. Water-based single-walled-nanotube-filled polymer composite with an exceptionally low percolation threshold. Adv Mater 2004;16(2):150–3. [25] Lovell PA, El-Aasser MS. In: Emulsion polymerization and emulsion polymers. New York: John Wiley & Sons, Inc.; 1997. [26] Yu CH, Shi L, Yao Z, Li DY, Majumdar A. Thermal conductance and thermopower of an individual single-wall carbon nanotube. Nano Lett 2005;5(9):1842–6. [27] Kirkpatrick S. Percolation and conduction. Rev Mod Phys 1973;45(4):574–88. [28] Deutscher G, Kapitulnik A, Rappaport M, Percolation structure and processes. In: Deutscher G, Zallen R, Adler J, editors. Annals of the israel physical society, vol. 5. New York: American institute of physics; 1983 [chapter 10]. [29] Vaisman L, Wagner HD, Marom G. The role of surfactants in dispersion of carbon nanotubes. Adv Colloid Interface 2006;128:37–46. [30] Bandyopadhyaya R, Nativ-Roth E, Regev O, Yerushalmi-Rozen R. Stabilization of individual carbon nanotubes in aqueous solutions. Nano Lett 2002;2(1):25–8. [31] Tummala NR, Striolo A. SDS Surfactants on carbon nanotubes: aggregate morphology. ACS Nano 2009;3(3):595–602. [32] Grossiord N, Loos J, Regev O, Koning CE. Toolbox for dispersing carbon nanotubes into polymers to get conductive nanocomposites. Chem Mater 2006;18(5):1089–99. [33] Wang H, Zhou W, Ho DL, Winey KI, Fischer JE, Glinka CJ, et al. Dispersing single-walled carbon nanotubes with surfactants: a small angle neutron scattering study. Nano Lett 2004;4(9):1789–93. [34] Moore VC, Strano MS, Haroz EH, Hauge RH, Smalley RE, Schmidt J, et al. Individually suspended single-walled carbon nanotubes in various surfactants. Nano Lett 2003;3(10):1379–82. [35] Grunlan JC, Liu L, Regev O. Weak polyelectrolyte control of carbon nanotube dispersion in water. J Colloid Interface Sci 2008;317(1):346–9. [36] Dror Y, Pyckhout-Hintzen W, Cohen Y. Conformation of polymers dispersing single-walled carbon nanotubes in water: a small-angle neutron scattering study. Macromolecules 2005;38(18):7828–36. [37] Ouyang BY, Chi CW, Chen FC, Xi QF, Yang Y. Highconductivity poly(3,4-ethylenedioxythiophene): poly(styrene sulfonate) film and its application in polymer optoelectronic devices. Adv Funct Mater 2005;15(2):203–8. [38] Ouyang J, Xu QF, Chu CW, Yang Y, Li G, Shinar J. On the mechanism of conductivity enhancement in poly(3,4ethylenedioxythiophene): poly(styrene sulfonate) film through solvent treatment. Polymer 2004;45(25):8443–50.

[39] Tsai YC, Chiu CC, Tsai MC, Wu JY, Tseng TF, Wu TM, et al. Dispersion of carbon nanotubes in low pH aqueous solutions by means of alumina-coated silica nanoparticles. Carbon 2007;45(14):2823–7. [40] Liu L, Grunlan JC. Clay assisted dispersion of carbon nanotubes in conductive epoxy nanocomposites. Adv Funct Mater 2007;17(14):2343–8. [41] Rahman GMA, Guldi DM, Campidelli S, Prato M. Electronically interacting single wall carbon nanotube-porphyrin nanohybrids. J Mater Chem 2006;16(1):62–5. [42] Li HP, Zhou B, Lin Y, Gu LR, Wang W, Fernando KAS, et al. Selective interactions of porphyrins with semiconducting single-walled carbon nanotubes. J Am Chem Soc 2004;126(4):1014–5. [43] Wenseleers W, Vlasov II, Goovaerts E, Obraztsova ED, Lobach AS, Bouwen A. Efficient isolation and solubilization of pristine single-walled nanotubes in bile salt micelles. Adv Funct Mater 2004;14(11):1105–12. [44] Haggenmueller R, Rahatekar SS, Fagan JA, Chun JH, Becker ML, Naik RR, et al. Comparison of the quality of aqueous dispersions of single wall carbon nanotubes using surfactants and biomolecules. Langmuir 2008;24(9):5070–8. [45] Fagan JA, Huh JY, Simpson JR, Blackburn JL, Holt JM, Larsen BA, et al. Separation of empty and water-filled single-wall carbon nanotubes. ACS Nano 2011;24(5):3943–53. [46] Chen JY, Collier CP. Noncovalent functionalization of singlewalled carbon nanotubes with water-soluble porphyrins. J Phys Chem B 2005;109(16):7605–9. [47] Hasobe T, Hattori S, Kamat PV, Fukuzumi S. Supramolecular nanostructured assemblies of different types of porphyrins with fullerene using TiO2 nanoparticles for light energy conversion. Tetrahedron 2006;62(9):1937–46. [48] Alvaro M, Atienzar P, la Cruz P, Delgado JL, Troiani V, Garcia H, et al. Synthesis, photochemistry, and electrochemistry of single-wall carbon nanotubes with pendent pyridyl groups and of their metal complexes with zinc porphyrin. Comparison with pyridyl-bearing fullerenes. J Am Chem Soc 2006;128(20):6626–35. [49] Chen G, Dresselhaus MS, Dresselhaus G, Fleurial JP, Caillat T. Recent developments in thermoelectric materials. Int Mater Rev 2003;48(1):45–66. [50] Asbeck WK, Vanloo M. Critical pigment volume relationships. Ind Eng Chem 1949;41(7):1470–5. [51] Lei M, Francis LF, Scriven LE. Correlation of hardness with MFFT and CPVC of latex/ceramic coatings. J Coating Technol 2003;75(945):95–102. [52] Weaver JC. CPVC, Critical pigment volume concentration – an overview. J Coating Technol 1992;64(806):45–6. [53] Grunlan JC, Gerberich WW, Francis LF. Electrical and mechanical behavior of carbon black-filled poly(vinyl acetate) latex-based composites. Polym Eng Sci 2001;41(11): 1947–62. [54] Grunlan JC, Kim YS, Ziaee S, Wei X, Abdel-Magid B, Tao K. Thermal and mechanical behavior of carbon-nanotube-filled latex. Macromol Mater Eng 2006;291(9):1035–43. [55] Ou YC, Yu ZZ, Vidal A, Donnet JB. Effects of alkylation of silicas on interfacial interaction and molecular motions between silicas and rubbers. J Appl Polym Sci 1996;59(8):1321–8. [56] Starr FW, Schroder TB, Glotzer SC. Molecular dynamics simulation of a polymer melt with a nanoscopic particle. Macromolecules 2002;35(11):4481–92. [57] Borodin O, Smith GD, Bandyopadhyaya R, Byutner E. Molecular dynamics study of the influence of solid interfaces on poly(ethylene oxide) structure and dynamics. Macromolecules 2003;36(20):7873–83. [58] Bansal A, Yang HC, Li CZ, Cho KW, Benicewicz BC, Schadler LS, et al. Quantitative equivalence between polymer

CARBON

[59]

[60]

[61]

[62] [63]

[64]

[65]

[66]

[67]

[68]

5 0 (2 0 1 2) 8 8 5–89 5

nanocomposites and thin polymer films. Nat Mater 2005;4(9):693–8. Grunlan JC, Grigorian A, Hamilton CB, Mehrabi AR. Effect of clay concentration on the oxygen permeability and optical properties of a modified poly(vinyl alcohol). J Appl Polym Sci 2004;93(3):1102–9. Benoit JM, Corraze B, Lefrant S, Blau WJ, Bernier P, Chauvet O. Transport properties of PMMA-carbon nanotubes composites. Synth Met 2001;121(1–3):1215–6. Berber S, Kwon YK, Tomanek D. Unusually high thermal conductivity of carbon nanotubes. Phys Rev Lett 2000;84(20):4613–6. Che JW, Cagin T, Goddard WA. Thermal conductivity of carbon nanotubes. Nanotechnology 2000;11(2):65–9. Prasher R. Thermal boundary resistance and thermal conductivity of multi-walled carbon nanotubes. Phys Rev B 2008;77(7):1–11. Yang DJ, Zhang Q, Chen G, Yoon SF, Ahn J, Wang SG, et al. Thermal conductivity of multi-walled carbon nanotubes. Phys Rev B 2002;66:165440–6. Yu CH. Nanomaterials characterization and bio-chemical sensing using microfabricated devices. Ph.D. Thesis. Austin Texas USA: University of Texas; 2004. Shi L, Li DY, Yu CH, Jang WY, Kim D, Majumdar A, et al. Measuring thermal and thermoelectric properties of onedimensional nanostructures using a microfabricated device. J Heat Trans-T Asme 2003;125(5):881–8. Zhong H, Lukes JR. Interfacial thermal resistance between carbon nanotubes: molecular dynamics simulations and analytical thermal modeling. Phys Rev B 2006;74: 125403–13. Pollack GL. Kapitza resistance. Rev Mod Phys 1969;41: 48–81.

895

[69] Huxtable S, Cahill D, Shenogin S, Xue L, Ozisik R, Barone P, et al. Interfacial heat flow in carbon nanotube suspensions. Nat Mater 2003;2(11):731–4. [70] Shenogin S, Liping X, Ozisik R, Keblinski P. Role of thermal boundary resistance on the heat flow in carbon-nanotube composites. J Appl Phys 2004;95(12):8136–44. [71] Yavari F, Fard HR, Pashayi K, Rafiee MA, Zamiri A, Yu ZZ, et al. Enhanced thermal conductivity in a nanostructured phase change composite due to low concentration graphene additives. J Phys Chem C 2011;115(17):8753–8. [72] Maeda T, Horie C. Phonon modes in single-walled nanotubes with a small diameter. Physica B 1999;263–264:479–81. [73] Kasuya A, Saito Y, Sasaki Y, Fukushima M, Maeda T, Horie C, et al. Size dependent characteristics of single wall carbon nanotubes. Mater Sci Eng A 1996;217–218:46–7. [74] Popov VN. Theoretical evidence for T1/2 specific heat behavior in carbon nanotube systems. Carbon 2004;42(5– 6):991–5. [75] Park YT, Ham AY, Grunlan JC. High electrical conductivity and transparency in deoxycholate-stabilized carbon nanotube thin films. J Phys Chem C 2010;114(14):6325–33. [76] Barros EB, Son HB, Samsonidze GG, Souza Filho AG, Saito R, Kim YA, et al. Double-wall carbon nanotubes treated with H2SO4. Phys Rev B 2007;76:045425. [77] D’yachkov PN, Makaev DV. Linear augmented cylindrical wave for calculating the electronic structure of double-wall carbon nanotubes. Phys Rev B 2006;74:155442. [78] Rao AM, Eklund PC, Bandow S, Thess A, Smalley RE. Evidence for charge transfer in doped carbon nanotube bundles from Raman scattering. Nature 1997;388:257–9. [79] Bandow S, Chen G, Sumanasekera GU, Gupta R, Yudasaka M, Iijima S, et al. Diameter-selective resonant Raman scattering in double-wall carbon nanotubes. Phys Rev B 2002;66:075416.