Multilayered Carbon Nanotube/Polymer Composite BasedThermoelectric FabricsCorey A. Hewitt,† Alan B. Kaiser,‡ Siegmar Roth,§ Matt Craps,∥ Richard Czerw,∥ and David L. Carroll* ,†
†Center for Nanotechnology, Wake Forest University, Winston Salem, North Carolina 27105, United States‡MacDiarmid Institute for Advanced Materials and Nanotechnology, SCPS, Victoria University of Wellington, Wellington 6140, New Zealand§School of Electrical Engineering, WCU Flexible Nanosystems, Korea University, Seoul, Korea∥NanoTechLabs, Yadkinville, North Carolina 27055, United States
*S Supporting Information
ABSTRACT: Thermoelectrics are materials capable of thesolid-state conversion between thermal and electrical energy.
Carbon nanotube/polymer composite thin films are known toexhibit thermoelectric effects, however, have a low figure of merit (ZT) of 0.02. In this work, we demonstrate individualcomposite films of multiwalled carbon nanotubes (MWNT)/polyvinylidene fluoride (PVDF) that are layered into multipleelement modules that resemble a felt fabric. The thermo-electric voltage generated by these fabrics is the sum of contributions from each layer, resulting in increased power output. Since these fabrics have the potential to be cheaper, lighter,and more easily processed than the commonly used thermoelectric bismuth telluride, the overall performance of the fabric showspromise as a realistic alternative in a number of applications such as portable lightweight electronics.
Traditional inorganic crystalline thermoelectrics such as
bismuth telluride (Bi2Te3) have been studied and utilizedcommercially for the last half century, but recent advancementsin organic thermoelectrics sho w promise for their use asalternatives to these materials.1 Organics typically have low electrical conductivities but they have the potential to be usedas thermoelectrics as a result of the inverse relationship
between the Seebeck coefficient α and electrical conductivity σ due to charge carrier concentration and mobility.2 Thesecompeting factors comprise the power factor α
2σ in the
dimensionless figure of merit (ZT = (α 2σ / κ )T where κ is thethermal conductivity and T is temperature) and is important
because it is directly related to the usable power attainable fromthe thermoelectric. A high ZT is achieved by creating a material
with a high power factor and low thermal conductivity. This
task is further complicated by the direct relationship betweenthe electrical conductivity and the charge carrier contribution tothe thermal conductivity.2 The figure of merit does not,however, include several other important considerations such ascost, weight, and processability. These additional factors allow for the consideration of organics as thermoelectric materials.
Of particular interest as an organic thermoelectric are carbonnanotube (CNT)/polymer thin films due to their heteroge-
neous structure that allows for the slight decoupling of these
thermoelectric parameters leading to an increased ZT.3−7
Currently the best nanotube/polymer thermoelectrics have aZT ≈ 0.02 and a power factor of 25 μ W m−1 K −2 , while Bi2Te3
has a ZT ≈ 1 and a power factor of 7 800 μ W m−1 K −2.6 At this
performance level, it would require a CNT/polymer thermo-electric module of about 500 cm2 to generate enough power torun a standard wrist watch from a ΔT ≈ 10 K generated by
body heat. This is about fifty times the area of a typical wrist watch. There are, however, several potential benefits to CNT/polymer thermoelectrics. Carbon nanotube/polymer compo-sites and Bi2Te3 have similar low thermal conductivities of about ∼3 W m−1 K −1.4,8 This allows for a sustainedtemperature difference across the film. The power per unitmass for Bi2Te3 is about 232 mWg−1 , while current CNTthermoelectrics have a power per unit mass of 60 mW g−1 ,9 buthave the potential to reach as high as 1300 mW g−1 if a ZT ≈
0.2 is reached. If CNT/polymer thermoelectrics are produced
on a large scale, the cost could be as low as $1/watt due to easeof production and low cost for materials, while currently produced Bi2Te3 thermoelectrics are ∼$7/watt.10 Additionally,CNT/polymer composites are flexible and durable, unlikecrystalline thermoelectrics. It is when these benefits areconsidered, along with ZT and the fact that it can be improvedupon, that the use of organics as thermoelectrics may bepractical in applications unsuited to Bi2Te3.
Received: October 28, 2011Revised: January 24, 2012Published: February 8, 2012
To produce sufficient power, any thermoelectric materialneeds to be combined into a module containing many alternating p-type and n-type legs that are connected electrically in series, and thermally in parallel. This arrangement of elements is utilized because it allows for the direct addition of the thermoelectric voltage contribution of each leg while it issubject to the same maximum available ΔT . This modulearchitecture results in the highest attainable thermoelectric
voltage for a given number of legs ( N ) subject to the availabletemperature difference ΔT . Typically, thermoelectric modulescomposed of bulk materials such as Bi2Te3 are arranged in a
way such that the temperature difference between the two endsof the thermoelectric legs is perpendicular to the surface of the
module.11
Since the CNT/polymer thermoelectric materials arethin films, however, this limits the maximum temperaturedifference attainable perpendicular to the surface of the film;therefore, a different geometry for connecting the legs must beadopted. In this Letter, we report on a method for constructinga thermoelectric module consisting of multiple layers of CNT/polymer films that allows for the arrangement of thetemperature gradient parallel to the surface of the module;this module arrangement results in a feltlike thermoelectricfabric.
The fabrication and characterization of the single films thatcomprise the multilayered fabric have been reportedpreviously.3 To form the multilayered film, individually prepared conducting and insulating layers are arranged as in
Figure 1a and then bonded together by pressing the stack at themelting point of the polymer in use (about 450 K). The filmsused in this study were polyvinylidene fluoride (PVDF) with 95or 20% CNTs by weight (wt %) for the conducting layers, andpure PVDF for the insulating layers. The resulting single filmthicknesses were 25−40 μm, while the multilayer film thicknessdepends on the total number of layers. The number of conduction layers is given by N = nn + np where nn and np arethe number of n-type and p-type layers, respectively. When thefabric is subject to a temperature difference ΔT = T h − T cparallel to the surface as shown in Figure 1 b, the charge carriers(holes h , or electrons e) travel from the T h side to the T c sidedue to the Seebeck effect and generate a thermoelectric voltage
V TEP (see Supporting Information Figure S1 for a description of how V TEP is affected by different temperature gradients). Sincethe voltage contribution of each conducting layer is determined
by its Seebeck coefficient and can be added in series due to thealternating p/n junctions, the resulting V TEP magnitude is given
by
= | | + | | ΔV n a n a T [ ]TEP n n p p (1)
where α n and α p are the Seebeck coefficients of the n-type andp-type films, respectively.
A room temperature measurement of V TEP/ΔT versus thenumber of conduction layers N for the 95 wt % fabrics wasperformed, with the results shown in Figure 2a. The theoretical
value is calculated from eq 1 using the room temperatureSeebeck coefficients of 10.05 μ V K −1 for α p and −5.04 μ V K −1
for α n. The experimental V TEP/ΔT values follow closely tothose calculated with no measurable drop off in V TEP as N isincreased. Adding layers to the fabric is equivalent to adding
voltage sources in series, so the limiting factor of N in practice
Figure 1. (a) Layer arrangement for the multilayered fabric. CNT/PVDF conduction layers (B,D) are alternated between PVDF insulation layers(A,C,E). Every other conduction layer contains p-type CNTs (B), while the others contain n-type CNTs (D). The shorter insulating layers allow foralternating p/n junctions when the stack is pressed and heated to the polymer melting point of 450 K to bond the layers. Layers A −D can berepeated to reach the desired number of conduction layers N. When the film is exposed to a temperature gradient ΔT , charge carriers (holes h , orelectrons e) migrate from T h to T c resulting in a thermoelectric current I . (b) The resulting thermoelectric voltage V TEP can be read across the endsof the first and last conduction layers. (c) The thermoelectric fabric remains flexible and lightweight.
Figure 2. (a) Thermoelectric voltage generated per 1 K ΔT versus thenumber of conduction layers N in a multilayered film composed of 95
wt % CNT/PVDF single films. The fit is calculated using eq 1 and theroom temperature Seebeck coefficients of α n = −5.04 μ V K −1 and α p =10.05 μ V K −1. (b) V TEP versus ΔT for a 72 layer fabric. T c was held atroom temperature while T h was increased to 390 K, at which point thefabric short circuited due to melting of the PVDF. The solid line showsthat V TEP increases linearly with a V TEP/ΔT value of 550 μ V K −1.
Nano Letters Letter
dx.doi.org/10.1021/nl203806q | Nano Lett. 2012, 12, 1307−13101308
8/23/2019 Carbon Nanotubes as Thermoelectric Chargers
is the heat source’s ability to produce a sufficient ΔT throughout all N layers. Since the V TEP is proportional to ΔT
because of the Seebeck effect, increasing ΔT will also result inan increased V TEP output, which is shown in Figure 2 b. A 72layer fabric was used with a V TEP/ΔT = 550 μ V K −1 leading to a51 mV output at a ΔT = 95 K. The limiting factor for ΔT is
when T h = 390 K because at this point the polymer beginsdeforming, compromising the multilayered structure of thefabric.
The temperature dependent behavior of V TEP/ΔT is alsoimportant because the fabrics will be subject to varyingtemperatures during use. The temperature dependence of V TEP/ΔT for the 20 wt % single films and several differentmultilayered fabrics was measured with the results shown inFigure 3a. The observed temperature dependent behavior of α
for the single films is typical of CNTs and CNT compositefilms.12−16 This trend has been described previously using aheterogeneous model given by
α = + −
+
⎜ ⎟
⎡
⎣⎢⎢ ⎛⎝ ⎞⎠
⎤
⎦⎥⎥T bT cT
T
T ( ) exp
d
1/2 0
1/ 1
(2)
where b and c are constants for the metallic and semi-conducting contributions, T 0 is a constant related to the energy differences for hopping between nanotubes, and d is thedimensionality of hopping that depends on the morphology of inter tube contacts.13−15 The T 1/2 term is exponentially
weighted to represent the suppression of the semiconductingcontribution to α at low T .16 The multilayered fits werecalculated using the single n-type layer and p-type layer fitsfrom eq 2 (to determine α n(T ) and α p(T )) along with thecorresponding nn and np values to calculate V TEP/ΔT using eq
1. The results show that the multilayered films retain thecharacteristic T dependence of the single films, while stillproducing the expected V TEP.
Electrical conductivity measurements were also performedon several multilayered fabrics to determine if layering the filmsin the module introduces any internal resistance to the overallfilm due to the p/n junctions. Figure 3 b shows the results of σ
versus absolute T for the N = 3, 7, 11 layer films, as well as thesingle film electrical conductivities. These results are typical of CNTs and CNT/polymer composites, and a previously reported thermal fluctuation induced tunneling model has
been used to describe the temperature dependent behav-ior.15 ,17,18 If no additional resistance is introduced due tolayering, the multilayered σ should be between the two singlefilm conductivities. From Figure 3 b, the electrical conductivity for N = 3 does fall between the single film values, but there isabout a 15% drop off in σ for the N = 7, 11 modules. For the N = 11 module, this resulted in an average decrease in σ of about1% per p/n junction.
The decrease in σ is most likely due to the decreased CNTconcentration in the p/n junction region. Figure 4a shows the
typical CNT arrangement and concentration for a 20 wt % film, while Figure 4 b shows the composition of the junction. Sincethe films are bonded together by pressing the junction at themelting point of the polymer, the main constituent of the
junction is polymer. The decrease in σ could potentially beeliminated by forming the single films in one continuous strip
with alternating p-type and n-type segments and then foldingthe alternating layers over, or by evaporating a highconductivity material such as indium ox ide onto the segmentof the film that will form the junction.19
Power measurements on the 72 layer fabric were performedfor several different load resistances with the results shown inFigure 5. The ΔT was kept at a safe operating temperature of 50 K to avoid deformation of the structure at high
temperatures. The maximum power generation of 137 nW occurred when the load resistance matched the internal fabricresistance of 1270 Ω. At this load resistance the V TEP was 13mV compared to an open circuit V TEP of 26 mV at the sameΔT . Above 1270 Ω , V TEP continues to increase as it approachesthe open circuit voltage, but P decreases as the load resistance
becomes exponentially larger. The power output as a functionof ΔT at a load resistance of 1270 Ω is shown in the inset of Figure 5 and exhibits a squared behavior due to the linearrelationship between V TEP and ΔT as seen in Figure 2 b.
If higher power levels are required, ΔT could be increased asshown in the Figure 5 inset, and the number of conductionlayers can be increased provided the heat source can supply asufficient ΔT . For a fabric composed of 300 layers and exposed
to a ΔT = 100 K, the theoretical power output could be as highas 5 μ W. Further improvements to power output couldpotentially be made through optimization of the single film ZT.This could be done by improving the Seebeck coefficientthrough chemical treatment of the nanotubes,20 increasingelectrical conductivity by using conducting polymers,1 ordecreasing thermal conductivity by introducing phononscattering defects along the nanotubes.21 With optimizationof the single film ZT, film dimensions, and multilayerinterfabric contacts and layer count for a specific application,these fabric-modules could offer a realistic alternative to currentthermoelectrics for use in lightweight, flexible, and portableelectronics.
Figure 3. (a) Temperature dependence of V TEP for the N = 3, 7, 11layer fabrics, along with α for the single 20 wt % CNT/PVDF p-typeand n-type films. Fits to single film data are calculated using eq 2 , whilemultilayer fits are calculated using the single film results coupled witheq 1. (b) Temperature dependence of electrical conductivity for the N = 3, 7, 11 layer fabrics and the single 20 wt % CNT/PVDF p-type andn-type films.
Nano Letters Letter
dx.doi.org/10.1021/nl203806q | Nano Lett. 2012, 12, 1307−13101309
8/23/2019 Carbon Nanotubes as Thermoelectric Chargers
The authors declare no competing financial interest.
■ ACKNOWLEDGMENTS
S. Roth acknowledges World Class University Project “FlexibleNanosystems” (WCU, R32-2008-000-10082-0) of the KoreanMinistry of Education, Science, and Technology.
■ REFERENCES
(1) Bubnova, O.; Khan, Z. U.; Malti, A.; Braun, S.; Fahlman, M.;Berggren, M.; Crispin, X. Nat. Mater. 2011 , 10 , 429−433.
(2) Snyder, G. J.; Toberer, E. S. Nat. Mater. 2008 , 7 , 106−114.(3) Hewitt, C. A.; Kaiser, A. B.; Roth, S.; Craps, M.; Czerw, R.;
Carroll, D. L. Appl. Phys. Lett. 2011 , 98 , 183110.
(12) Baxendale, M.; Lim, K. G.; Amaratunga, G. L. Phys. Rev. B 2000 ,61 (19), 12705−12708.(13) Choi, Y.-M.; Lee, D.-S.; Czerw, R.; Chiu, P.-W.; Grobert, N.;
Terrones, M.; Reyes-Reyes, M.; Terrones, H.; Charlier, J.-C.; Ajayan,P. M.; Roth, S.; Carroll, D. L.; Park, Y.-W. Nano Lett. 2003 , 3 (6),839−842.
(14) Carroll, D. L.; Czerw, R.; Webster, S. Synth. Met. 2005 , 155 ,694−697.
(15) Kaiser, A. B.; Du s̈berg, G.; Roth, S. Phys. Rev. B 1998 , 57 (3),1418−1421.
(16) Kaiser, A. B.; Park, Y. W.; Kim, G. T.; Choi, E. S.; Du s̈berg, G.;Roth, S. Synth. Met. 1999 , 103 , 2547−2550.
(17) Kymakis, E.; Amaratunga, G. A. J. Appl. Phys. 2006 , 99 , 084302.(18) Sheng, P. Phys. Rev. B 1980 , 21 (6), 2180−2195.(19) Tahar, R. B. H.; Ban, T.; Ohya, Y.; Takahashi, Y. J. Appl. Phys.
1998 , 83 (5), 2631−2645.
(20) Collins, P. G.; Bradley, K.; Ishigami, M.; Zettl, A. Science 2000 ,287 , 1801−1804.
(21) Che, J.; C a̧g ı̆n, T.; W., A. G. III Nanotechnology 2000 , 11 , 65−69.
Figure 4. (a) SEM image of the surface of one of the 20 wt % CNT p-type legs. The CNT/polymer matrix is visible with the ∼30 nm diameter tubescomprising most of the surface while the polymer is coating the CNTs and binding them together. (b) Image of the p/n junction (between greendashed lines), which was formed by heating the junction region to slightly above the polymer melting point (450 K). Nanotubes are visible in the junction, but the main constituent is polymer (brighter regions), as opposed to the surface composition in (a).
Figure 5. Thermoelectric power and voltage generated by a 72 layerfilm at a ΔT = 50 K for varying load resistances. The peak power of 137 nW occurs at a load resistance of 1270 Ω. The inset shows power versus ΔT for a load resistance of 1270 Ω. This squared behavior wasexpected due to the linear trend of V TEP versus ΔT in Figure 2 b andthe relation P = V 2/R .
Nano Letters Letter
dx.doi.org/10.1021/nl203806q | Nano Lett. 2012, 12, 1307−13101310
