Continuous Carbon Nanotube-Based Fibers and Films forApplications Requiring Enhanced Heat DissipationPeng Liu,†,‡ Zeng Fan,‡ Anastasiia Mikhalchan,‡ Thang Q. Tran,‡ Daniel Jewell,§ Hai M. Duong,*,‡

and Amy M. Marconnet*,†

†School of Mechanical Engineering and the Birck Nanotechnology Center, Purdue University, West Lafayette, Indiana 47907, UnitedStates‡Department of Mechanical Engineering, National University of Singapore, 9 Engineering Drive 1, EA-07-05, Singapore 117575,Singapore§Department of Materials Science and Metallurgy, University of Cambridge, Cambridge CB2 1TN, United Kingdom

*S Supporting Information

ABSTRACT: The production of continuous carbon nanotube (CNT) fibers and films has paved the way to leverage thesuperior properties of individual carbon nanotubes for novel macroscale applications such as electronic cables and multifunctionalcomposites. In this manuscript, we synthesize fibers and films from CNT aerogels that are continuously grown by floating catalystchemical vapor deposition (FCCVD) and measure thermal conductivity and natural convective heat transfer coefficient from thefiber and film. To probe the mechanisms of heat transfer, we develop a new, robust, steady-state thermal characterizationtechnique that enables measurement of the intrinsic fiber thermal conductivity and the convective heat transfer coefficient fromthe fiber to the surrounding air. The thermal conductivity of the as-prepared fiber ranges from 4.7 ± 0.3 to 28.0 ± 2.4 W m−1 K−1

and depends on fiber volume fraction and diameter. A simple nitric acid treatment increases the thermal conductivity by as muchas a factor of ∼3 for the fibers and ∼6.7 for the thin films. These acid-treated CNT materials demonstrate specific thermalconductivities significantly higher than common metals with the same absolute thermal conductivity, which means they arecomparatively lightweight, thermally conductive fibers and films. Beyond thermal conductivity, the acid treatment enhanceselectrical conductivity by a factor of ∼2.3. Further, the measured convective heat transfer coefficients range from 25 to 200 Wm−2 K−1 for all fibers, which is higher than expected for macroscale materials and demonstrates the impact of the nanoscale CNTfeatures on convective heat losses from the fibers. The measured thermal and electrical performance demonstrates the promisefor using these fibers and films in macroscale applications requiring effective heat dissipation.

KEYWORDS: carbon nanotubes, thermal conductivity, convective heat transfer, CNT macroscale assemblies, acid treatment

1. INTRODUCTION

Individual carbon nanotubes (CNTs) demonstrate excellentmechanical (Young’s modulus up to 1 TPa),1 electrical (up to106 S cm−1),2 and thermal (up to 3500 W m−1 K−1)3

properties. Many applications could benefit from CNT-basedmaterials, including high-strength and conductive nanocompo-sites, electrochemical devices, sensors, and probes.4−6 However,practical applications require the extension of the superiorproperties of individual CNTs up to the macroscale. Thisremains a challenge due to the difficulty in synthesizing

homogeneous and uniform quality CNT-based structures thatfully exploit the excellent axial properties of individual CNTs.Manufacturing CNTs into continuous, macroscale fibers is a

significant step toward integrating them into practicalapplications. CNT fibers, produced by either wet-spinningfrom CNT/acid or polymer solutions,7,8 dry-spinning from

Received: April 6, 2016Accepted: June 20, 2016

Research Article

www.acsami.org

© XXXX American Chemical Society A DOI: 10.1021/acsami.6b04114ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

vertically aligned CNT arrays,9 or direct-spinning from CNTaerogels formed in floating catalyst chemical vapor deposition(FCCVD)10−15 are promising as novel lightweight, high-strength, and highly conductive materials. Compared withcommercial carbon and polymer fibers, CNT fibers demon-strate higher modulus and strength, better electrical andthermal conductivities, and more flexibility.8,14−16 The effectsof fabrication processes17,18 and different post-treatmentmethods8,14,16,19,20 on the morphology, mechanical, andelectrical properties have been systematically studied for severalyears. But there are only a few reported experimental studies onthe thermal properties of CNT fibers,8,21−27 and the largedifference in reported thermal conductivities between thesesimilar fibers indicates a significant dependence on quality andprocessing technique.CNT fibers can be assembled directly from CNT aerogels in

FCCVD10,15 leveraging the ability to produce CNT fibers inlarge scale efficiently with controllable CNT structures andoutstanding mechanical and electrical properties.10,14,15 This isin contrast to the wet-spinning and array-spinning methodswhere individual CNTs are first produced as powders or arraysand require a separate postprocess of spinning to create thefibers. Although the thermal conductivity of the CNT fibersdirectly spun from FCCVD-synthesized CNT aerogels has notyet been reported in detail, they are expected to have highthermal conductivity.5,28 Despite years of research, the reportedeffective thermal conductivities of CNT fibers are still muchlower than those of individual CNTs. High defect concen-tration and high thermal contact resistance between CNTs inmacroscale assemblies are commonly cited reasons for the lowperformance of CNT-based networks.29 Chemical modifica-tions to the CNTs, such as acid treatments, have beendemonstrated to improve CNT quality and manage theinterfacial interactions.30 For instance, soaking CNT fibers innitric acid increases their electrical conductivity.20,31 But littlethermal transport data exists for such modifications, and theeffect of acid treatment on the thermal conductivity of CNTmacroscale assemblies has not yet been experimentallyinvestigated.In this work, we measure the thermal conductivity and heat

transfer coefficient by natural convection of CNT fibers andthin films fabricated from CNT aerogels, which are grown byFCCVD. We develop a steady-state Joule heating infraredthermal metrology technique to study the thermal conductivityand convective heat transfer for the CNT-based fibers. Thistechnique leverages infrared microscopy as a noncontact,nondestructive, fast thermal characterization tool to measurethe two-dimensional temperature maps in the CNT fibers andfilms self-heated by Joule heating. As opposed to conventionalmeasurement techniques, the sample thermal conductivity canbe extracted in air and without need for microfabrication.Further, the steady-state technique developed here enables adirect measurement of thermal conductivity (independent ofdensity and specific heat) while simultaneously measuringconvective heat transfer coefficient. The technique is applicableto a broad class of electrically conductive samples ranging frommetallic wires to CNT fibers. Here, the impact of micro/macrostructure (collecting time, diameter, and density) on thethermal conductivity and the convection coefficient for theCNT fibers is investigated for the first time. These datacomplement previous work characterizing the mechanical andelectrical properties of the CNT fibers directly spun from CNTaerogels in CVD.13,14 Moreover, we demonstrate a simple,

effective acid treatment approach to enhance the electrical andthermal conductivities of the CNT fibers and films. The acid-treated specimens demonstrate up to a factor of ∼3 for the fiberand ∼6.7 for the thin film improvement in the thermalconductivities than the as-prepared samples due to thepurification and densification during the acid treatment,which results in less defective structure and reducing contactresistance between nanotubes.

2. EXPERIMENTAL SECTIONFabrication of CNT Fibers and Films. The CNT fibers are

fabricated from CNT films, which are synthesized by the floatingcatalyst vapor deposition method.13 In a reducing hydrogenatmosphere, the nanotubes form an aerogel in the hot zone of thefurnace (1200 °C) and the aerogel is stretched into cylindrical hollowsocks. The CNT sock is pulled out of the furnace by a stainless rodand continuously collected on a paper roller to form a CNT film. CNTfilms collected for 1, 5, and 10 min are mechanically rolled from oneend to the other to fabricate the CNT fibers with various diametersand densities.13 In order to study the effects of acid treatment on thethermal conductivity, the CNT fibers and films are immersed in 65 wt% nitric acid (HNO3, Sigma-Aldrich Pte Ltd.) at room temperature for30 min and washed in deionized water.

Morphology Characterizations. The structures of the specimensare observed using a field emission scanning electron microscope (FE-SEM, Model S-4300, Hitachi), a transmission electron microscope(TEM, JEOL JEM-3010), and a Horiba Jobin Yvon Modular RamanSpectrometer. Nitrogen adsorption/desorption tests to determine thedistribution of pores in the specimens before and after acid treatment,are conducted using a Nova 2200e Surface Area and Pore SizeAnalyzer (Quantachrome). To measure the bulk density of thespecimens, the masses of the specimens are obtained with an analyticalbalance in accuracy of 0.01 mg and the dimensions of the specimensare measured by an Olympus optical microscope. The CNT volumefraction ( f) is calculated from the bulk density of the CNT fiber orfilm, ρ, as

ρρ

= ×f 100%CNT (1)

where ρCNT = 2.1 g cm−3 is the density of an individual CNT.Electrical Properties of the CNT Fibers and Films. Electrical

resistances of the CNT fibers and films are measured by four-wiremethod using a Keithley 2420 source meter. For better electricalcontacts, the two ends of the specimens are fixed on two separate glassslides with silver paint. The electrical conductivity (σ) and specificelectrical conductivity (σ′) is then calculated from eqs 2 and 3,respectively,

σ = −LRA

(S cm )1(2)

σ σρ

′ = −(S cm g )2 1

(3)

where R, A, and L are the resistance, cross-sectional area, and length ofthe specimen.

The temperature coefficient of resistance (TCR or α) is measuredby four-wire electrical resistance measurements across the temperaturerange of 295−350 K. The average TCR of the as-prepared specimensand acid-treated specimens was 0.001 15 ± 0.000 25 and 0.002 65 ±0.000 40 K−1, respectively.

Thermal Properties of the CNT Fibers and Films. Each CNTfiber is suspended between two pairs of copper blocks working aselectrodes as shown in Figure 1a. For the CNT fiber samples, thelength of suspended portion is ∼1 cm; while for the film samples, thesuspended portion is 2 cm in length and 0.5 cm in width. Currentflowing through the sample heats the specimen. A Keithley 2420source meter (Figure 1b) is used to source the current and measurethe voltage drop across the sample. The temperature profile along the

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.6b04114ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

B

specimen is measured by an infrared temperature measurementmicroscope system (InfraScope, Quantum Focus InstrumentsCorporation), as shown in Figure 1c. The high spatial resolution(11.7 μm/pixel at 1× magnification) of the infrared microscopeenables hundreds of temperature data points along the length of a 1cm long specimen, as shown in Figure 1d. The one-dimensional (1D)heat transfer along the length of the fiber is confirmed by the nearlyuniform temperature in the direction perpendicular to the axis of thefiber as illustrated in Figure 1c and d and is further validated in detailin the Supporting Information. All measurements are performed underambient environment at room temperature (295.15 K). The range ofcurrents tested is limited to ensure the average temperatures of thespecimens do not exceed 303.5 K (near room temperature). Theemissivity of the samples is 0.86 ± 0.02 measured through a calibrationprocedure described in the Supporting Information.Heat transfer along the specimen is governed by the steady-state 1D

heat diffusion equation:32,33

α+ ′ + − − ′ − =kAT xx

p T x T h T x Td ( )

d[1 ( ( ) )] ( ( ) ) 0

2

2 0 0

(4)

with the boundary condition of

= ± =⎜ ⎟⎛⎝

⎞⎠T x

LT

2 L/2 (5)

where k, A, L, and α are the thermal conductivity, cross-sectional area,length, and TCR of the specimen, respectively, h′ is the convectiveheat loss per unit length to the air (with units of W m−1 K−1), T(x) isthe temperature profile along the length of the specimen, T0 is theroom temperature (295.15 K), TL/2 is the temperature of the ends ofthe specimen, p′ = I2R0/L is the heat generation per unit length in thespecimen, and I and R0 are current amplitude and the electricalresistance of the specimen at T0, respectively. The analytic solution ofeq 4 is

α= −

′′ − ′

− + −⎛

⎝⎜⎜

⎞

⎠⎟⎟( )

( )

T x Tp

h pmx

T T

mx

( )cosh( )

cosh1 ( )

cosh( )

cosh

mL L

mL

0

2

/2 0

2 (6)

where m2 = (h′ − αp′)/kA, when h′ − αp′ > 0. We fit eq 6 to theexperimental temperature profiles to measure the thermal conductivityk and convective heat transfer coefficient h. Figure 1d shows anexample of the best fit curve with an example experimentaltemperature profile. The traditional convection coefficient (h inWatts per squared meter per Kelvin) was then calculated from

= ′h h P/ (7)

where P is the perimeter of the cross section of the specimen.Note that the analytical solution in eq 6 assumes a constant

convective coefficient along the length wire. As the maximumtemperature of the wire is limited to a few Kelvin above the setbase temperature, the variation in the natural convection coefficient issmall. Essentially, the extracted heat transfer coefficient can beconsidered an average value for the fiber. This assumption is validatedin detail in the Supporting Information.

The proposed method was first verified by measuring the thermalconductivity of the bare nickel chromium resistance (BNC) wires(AWG 30, Consolidated Electronic Wire & Cable, United States). Asshown in Table 1, all experimental values are very close to the valuesprovided by the manufacturer, with standard deviations of less than10%. Thus, the experimental setup and methods employed in thepresent work were considered to be valid and extended to the CNTfiber and film samples.

3. RESULTS AND DISCUSSION3.1. Electrical Properties of CNT Fibers. Figure 2 shows

the effect of the volume fraction of CNTs on the electricalconductivity determined according to eq 2. When the volumefraction increases from 5.5% to 41.4%, the electricalconductivities of the fibers increases from 150 to 1050 Scm−1. While there is a clear increase in the electricalconductivity with increasing volume fraction, the specificelectrical conductivity, defined as the ratio between fiberelectrical conductivity and bulk density, is nearly independentof the volume fraction varying in the range of 1000−1500 Scm2 g−1. The average specific electrical conductivity is ∼1200 Scm2 g−1 obtained by linearly fitting the data for electricalconductivity as a function of fiber bulk density. A similar trendof increasing electrical conductivity with volume fraction, butconstant specific electrical conductivity, was previouslyobserved by Miao17 for array-spun CNT fibers. Here, themeasured specific electrical conductivity is comparable with

Figure 1. Experimental setup and example data from the steady stateJoule heating infrared metrology technique. (a) Optical image of aCNT fiber suspended between two copper electrodes that also serve asheat sinks. (b) Schematic of the experimental setup. The Keithley2420 source meter sources current through the sample leading to anincreased temperature throughout the fiber, which is measured by theinfrared microscope. (c) Example infrared microscope image of a CNTfiber during Joule heating for a sample with a measured density of 0.28g cm−3. Note the one-dimensional heat transfer along the length of thefiber as indicated by the nearly uniform temperature in the directionperpendicular to the axis of the fiber. (d) Temperature profile alongthe axis of the fiber. Solid lines indicate temperature measured alongthree selected rows of pixels between the center and edge of the fiber.Blue crosses indicate data points averaged from each column of pixelsin the 2D image. The bold solid red line indicates the best fit to theexperimental temperature profile with an extracted thermal con-ductivity of 9.5 W m−1 K−1 and convection coefficient of 37 W m−2

K−1.

Table 1. Data Sheet and Experimentally Measured Properties of Bare Nickel Chromium Resistance (BNC) Wires

diameter (μm) resistance (Ω cm−1) TCR (K−1) thermal conductivity (W m−1 K−1)

value provided by manufacturer 254 0.2215 0.00015 13experimental value 255 ± 5 0.2195 ± 0.005 0.000146 ± 0.00002 13 ± 1

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.6b04114ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

C

previously published result on CNT fibers synthesized fromaerogels (1600 S cm2 g−1)34 and one measurement of CNTfibers synthesized from aligned CNT arrays (1300−1400 S cm2

g−1).35 But it is much higher than the specific electricalconductivities of several other studies of CNT fiberssynthesized from aligned CNT arrays (370−900 S cm2

g−1).17,21,22

The average temperature coefficient of resistance (TCR)near room temperature for the CNT fibers is 0.00115 ±0.00025 K−1. Reported values of the TCR of CNT fibers nearroom temperature vary from 0.001 to 0.002 K−1.8,9,19,21 Notethat the TCR of the CNT fibers in this research isapproximately half that of the TCR of copper (∼0.004 K−1)and aluminum (∼0.0043 K−1), which is a promising result forfurther developing these CNT fibers into potential applications,such as electrical cables.5

3.2. Thermal Conductivities of CNT Fibers. In this work,thermal conductivities of CNT fibers with different collectingtime (1, 5, and 10 min) and volume fractions (5.5%−41.4%)are measured with a steady state Joule heating infraredmetrology technique. First, in order to investigate the effectof collecting time and diameter, fibers with the same volumefraction (∼17%) are made from 1, 5, and 10 min-films. Asshown in Figure 3a, the diameters of the fibers increase with theincreasing collecting time. Larger fiber volume is expectedbecause the longer collecting time leads to increasedaccumulated CNT mass while the volume fraction is keptconstant. The thermal conductivities of the fibers made from 1,

5, and 10 min-films are 13 ± 1.2, 12.5 ± 1.5, and 13.7 ± 2.0 Wm−1 K−1, respectively, which are nearly independent to thediameter and collecting time.In contrast to diameter and collecting time, the volume

fraction of CNTs has a great influence on the fiber thermalconductivity. Figure 3b shows the thermal conductivities of theCNT fibers as a function of volume fraction. When the volumefraction of CNT increases from 5.5% to 41.4%, the thermalconductivities of the fibers increase from 4.7 ± 0.3 to 28.0 ± 2.4W m−1 K−1. Similar to the specific electrical conductivity, thespecific thermal conductivity (k′) is also nearly independent ofthe volume fraction as illustrated by the nearly linearrelationship between the volume fraction and the thermalconductivity. The average specific thermal conductivity isestimated to be 32 W m2 kg−1 K−1 by linearly fitting thethermal conductivity as a function of the fiber bulk density.Heat conduction in carbon materials is usually dominated by

phonons due to the strong covalent sp2 bonding resulting inefficient heat transfer by lattice vibrations. Here, we predict thatthe electronic contribution to the thermal conductivity (ke) isless than 5% of the total measured conductivity for all thespecimens, as calculated via Wiedemann−Franz law: ke/σ = LT,where σ, L, and T are electrical conductivity, Lorenz number,and temperature, respectively.36

The CNT fiber can be regarded as a two-phase systemconsisting CNTs and air with thermal conductivities of kCNTand kair (0.0259 W m−1 K−129). A simple model consideringnanotubes conducting heat parallel with the air predicts that thethermal conductivity increases linearly with volume fraction ofCNTs ( f):6

= + − ≈k fk f k fk(1 )specimen CNT air CNT (8)

A linear fit to the data in Figure 3b provides an estimate of 68W m−1 K−1 for the thermal conductivity of the individualnanotubes in the fibers. This is within the reasonable range forindividual MWNTs. Previous researchers have reported 34 Wm−1 K−1 for the MWNT with a diameter of 14 nm and 160 Wm−1 K−1 for the MWNT with a diameter of 11.4 nm.38

However, it is much lower than other previous measurementsof individual MWNTs.39−44

Previous research suggests that the density of nanotubes isnot the only factor influencing the thermal conductivity ofnanotube-based systems. The low apparent thermal con-ductivity of the CNT fibers compared to that predicted fromthe high values of thermal conductivity of the individual CNTs

Figure 2. Electrical conductivity and specific electrical conductivity ofthe CNT fibers as a function of volume fraction. Note that theelectrical conductivity increases approximately linear with volumefraction indicating a nearly constant specific electrical conductivity.

Figure 3. Impact of (a) collecting time and (b) volume fraction on thermal conductivity. The thermal conductivity of the fibers is independent ofcollecting time and fiber diameter but increases approximately linearly (blue dotted line) with volume fraction indicating an effect individual CNTthermal conductivity of ∼68 W m−1 K−1. The green solid line indicates the best fit of the model of Chalopin et al.,37 which enables estimation of theCNT−CNT contact conductance of GCNT−CNT ∼ 13−130 pW K−1.

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.6b04114ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

D

reported in literature can be explained by the impact ofcontacts, as well as the quality and morphology of the CNTfibers.Thermal contact resistance leads to temperature jumps at the

interfaces between contacting nanotubes. The contact pointsmay also introduce additional phonon scattering or dampingphonon modes within a nanotube reducing the effectivethermal conductivity of each nanotube.6 Although the couplingin nanotube bundles significantly increases the contact area, itmay also lead to the deformation of nanotube shape due to vander Waals force between adjacent nanotubes.45 The deforma-tion could lead to suppression of optical phonon modes42 andincreased inter-CNT contact resistance,46 which results insubstantially reduced thermal transport abilities for theindividual CNT.The CNT−CNT contact conductance (GCNT−CNT) in this

work is estimated to be on the order of 13−130 pW K−1,extracted by fitting the model of Chalopin et al.37 to ourexperimental results (see Figure 3b):

π

ρ

ρ≈ −k

LD

G0.182specimen

specimen

grapheneCNT CNT

(9)

where ρgraphene = 7.6 × 10−7 kg m−2 is the surface mass densityof graphene. Note that this model was developed for randomlyoriented nanotube mats instead of aligned CNT films or fibers,and neglects the finite thermal conductivity of the individualCNTs. But, essentially, this value can be considered an upperbound to the thermal resistance for the contacts and it iscomparable to the few reported experimental and simulationworks (ranging from ∼3−50 to >100 pW K−1).37,47,48

Mesoscopic models for aligned CNT macroscale assemblies,which can fully model the interactions between the individualCNT properties, specimen morphology including CNT length,diameter and alignment and CNT−CNT contacts on thethermal conductivity, are required in the future to betterunderstand the heat conduction in CNT macroscale assemblies.In addition to contact resistances, thermal conduction within

nanotube fibers and films is complicated by CNT quality andthe morphology of the nanotubes. MWNTs grown by CVD aregenerally more defective than those grown by arc-discharge orlaser ablation methods.49 Simulations and measurements ofindividual CNTs with defects showed that the thermalconductivity decreased significantly with increasing defectconcentration. CNT fibers or films consisting of longernanotubes may exhibit even lower conductivity than the oneconsists of shorter nanotubes because of higher probability of adefect occurring within a longer tube.6 Additionally, thestructural inhomogeneity of the CNT macroscale assembliesstrongly impacts their apparent thermal conductivity.21,42 Asshown in Figure 4, even though CNTs are assembled intobundles with high orientation preference (Figure 4a), we stillobserve dangling ends and interweaving in the structure (Figure4b). These inhomogeneities may prevent the overall thermalconductivity from achieving the performance expected fromthat of individual CNTs.3.3. Convective Heat Transfer in CNT Fibers. Natural

convection from the CNT fibers is also investigated in thissame experiment. The physics of solid to gas heat transferchanges considerably with size50 and environmental factors liketemperature and pressure.51 Since most microelectronic andmicroelectro-mechanical devices operate in air, understandingthe mechanism of convective heat transfer at microscale is

critical for accurate predictions of temperature and deviceperformance.Figure 5a shows the convection coefficients of the CNT

fibers with various collecting times (tc) and diameters vary from26 to 193 W m−2 K−1, which is higher than the typical valuesfor the natural convection coefficient at the macroscale (2−25W m−2 K−1). In general, most thermal conductivity measure-ments of fibers require a high vacuum environment to eliminateconvection losses, which increases the experimental complexity.As a result, there are fewer studies (see Figure 5b) on themechanism of convective heat transfer at microscale than at themacroscale. Li et al.52 reported convection coefficient of 1039−1143 W m−2 K−1 for CNT fibers an order of magnitude smallerin diameter than our fibers (d ∼ 36−43 μm compared to∼150−800 μm in this work). Hsu et al.53 found the heattransfer coefficient between an individual single-walled nano-tube (SWNT) and the surrounding air molecules ranged from1.5 × 103 to 7.9 × 104 W m−2 K−1. Other studies on freeconvection from other microscale materials showed that theconvection coefficient varies from 100 to 7000 W m−2 K−1

depending on the diameter of the microwire.50−52,54−56 Ourobservation is consistent with these results in the literature,suggesting that the convection coefficient for CNT fibers is alsostrongly influenced by scaling effects, as well as other structuraland environmental factors.As shown in Figure 5a, for fibers with the same collecting

time (e.g., 10 min), the convection coefficient decreases withincreasing diameter. This trend is consistent with the resultscalculated by the correlation of Churchill and Chu29 for freeconvection on long horizontal cylinders:

= ++

≲

⎧⎨⎪⎪

⎩⎪⎪ ⎡

⎣⎢⎤⎦⎥

⎫⎬⎪⎪

⎭⎪⎪( )

NuRa

Ra

0.600.387

1

for 10

DD

1/6

0.559Pr

9/16 8/27

2

D12

(10)

where =Nu hD k/D air, RaD, and Pr are the Nusselt number,Rayleigh number, and Prandtl number, respectively. Note thatthe correlation of Churchill and Chu is for free convection at aconstant temperature, while the temperature distribution alongthe CNT fiber shows a cosh-shape. Since the averagetemperatures of the specimens were controlled within a small

Figure 4. SEM images illustrating (a) the overall alignment of CNTswithin the in the fibers and (b) the inhomogeneity of the structure thatmay depress the measure valued of thermal conductivity.

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.6b04114ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

E

range (296.15−303.15 K), no significant impact of spatiallyvarying convection coefficient is expected (see Figure S3 in theSupporting Information). Thus, a typical average fiber temper-ature (299.15 K) is chosen for use in the Churchill−Chucorrelation. As shown in Figure 5a, most of our experimentalresults are larger than the corresponded calculated values,although the diameter scaling effect has been taken intoaccount in this correlation.In addition, the correlation of Churchill and Chu29 indicates

that the convection coefficient monotonically decreases withthe increasing diameter, independent of other fiber properties.However, the convection coefficients of the CNT fibers inFigure 5a are also affected by the collecting time, whichdetermines the number of CNTs inside the fiber. This suggeststhat the porous structure of the CNT fibers may affect theconvective heat transfer considerably. Since it is difficult tocount the number of individual CNTs inside a fiber, here weinvestigate the effect of the volume fraction, reflecting thenumber of individual CNTs per unit volume. As shown inFigure 5c, the convection coefficient increases with increasingof volume fraction. Previous studies report that surface area tovolume ratio demonstrates great impact on heat transfercoefficient.51,57 Unlike solid materials, the effective surface areaof CNT fibers is difficult to calculate. The insets in Figure 5cshow illustrations of the cross-section of a CNT fiber. Thestructure of the CNT fibers can be regarded as numerous CNTbundles consisting of aligned individual CNTs. Open pores,between the CNT bundles, interconnect the air inside andoutside of the fiber. Thus, the free convection can beconsidered on not only the outer surface but also the surfaceof each or some CNT bundles inside. The convectioncoefficient reflects the degree of heat loss per unit area.

Assuming a fixed outer diameter and a fixed convectioncoefficient on CNT bundle surface, a fiber with higher volumefraction consists of more CNT bundles and thus could have alarger effective surface area for convection. This results in moretotal heat loss and a higher effective convection coefficientbecause only the outer perimeter (P = πD) is used to calculateh in eq 7. For a better understanding of the convective heattransfer in a porous structure like the CNT fiber here, wesuggest to consider an effective diameter (Deff (m)), whichcombines both the outer diameter (D (m)) and volumefraction ( f (%)) by the following empirically obtained equation:

= +− −D D D f(0.000268 0.061)eff1.6 1.8

(11)

As shown in Figure 5c and d, the corresponding convectioncoefficients calculated by eqs 10 and 11 have a goodconsistency with the experimental data. To a certain extent,this estimate of effective diameter indicates synergistic effectsbetween the outer diameter and the number of CNT inside thefiber on the convective heat transfer in these porous CNTfibers. The actual convection between CNT fibers and the air iscomplicated due to the interaction of many environmentalvariables, as well as the complex structure of the CNT fiber.Ultimately, a more detailed correlation may include manyfactors including CNT length, bundle size, bundle entangle-ment, pore size, and gas flow inside the pore structure.The mechanism of heat transfer between microscale

materials and air or other gases is also controversial. Someresearch assumes that the dominant mode of heat transfer atthe microscale is conduction through the air, instead of theclassic notion of advection, such that natural convection effectscould be neglected at microscale.57 Models considering the

Figure 5. (a) Impact of diameter and collecting time on the convection heat transfer coefficient. The trend of decreasing convection coefficient withdiameter agrees well with the correlation of Churchill and Chu (calculated at 299.15 K),29 (dashed line), although exact values differ. Note that forthe same diameter, a higher collecting time indicates more CNTs within the same outer fiber diameter. (b) Measured convection coefficients of thiswork compared with literature data52,55,56 and predictions50 for microwires. Note that the convection coefficient for CNT fibers is not only stronglyinfluenced by scaling effects but also by other structural and environmental factors. (c) Effect of volume fraction on the convection coefficient.(inset) Cross-section of a CNT fiber with different volume fractions. (d) Convection coefficient of the CNT fiber as a function of effective diameter.The red hollow circles in part c and the solid red line in part d indicate the calculated convection coefficient using the Churchill−Chu correlation andthe effective diameter of the fiber.

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.6b04114ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

F

scaling effect and the microstructure of the CNT fiber arerequired in the future to further study the mechanism of heattransfer in CNT based assemblies.3.4. Effects of Acid Treatment on Thermal Con-

ductivity. To study the effects of acid treatment on thethermal conductivity, both CNT fibers and the CNT films fromwhich they were fabricated are immersed in 65 wt % HNO3 atroom temperature for 30 min. The effects of acid treatment onphysical properties of CNT fiber and films are summarized inTable 2. After acid treatment, the electrical conductivities of the

specimens are enhanced by a factor of 2−3, while their thermalconductivities are enhanced by a factor of 3−6. Additionally,the specific electrical and thermal conductivities are enhancedby factors of ∼2 and ∼4, respectively. For the as fabricatedsamples discussed previously, the specific electrical and thermalconductivities are nearly independent of density. However, acidtreatment significantly improves the specific conductivitiesindicating that the intrinsic conductivity of the CNTs or theinterfacial resistance have been modified during the acidtreatment.Such significant improvements on both electrical and thermal

conductivities of the specimens can be related to structuralmodifications, which are investigated using Raman spectrosco-py based on the ratio ID/IG. As shown in Figure 6a, ID/IGdecreases from 0.26 to 0.21 after acid treatment, suggesting thatCNTs become less defective after this treatment. Chemicalprocessing of CNTs by concentrated nitric acid induces severalmodifications to the structure including purification (i.e., toeliminate amorphous carbon adhering to the CNTs),functionalization, and decapping of the CNTs.30 Functionaliza-tion typically requires elevated temperature or long times (90−120 min or more),20,30,31 which may result in an increase in theratio ID/IG after acid treatment.20,31 Unlike those studies, in thiswork, the CNT fibers and films are treated for just 30 min atroom temperature and purification is expected to be thedominant effect in this initial short period of acid treatment.The impact of the nitric acid is further elucidated by TEMimages, as shown in Figure 6c and d. Before the acid treatment,many small particles could be observed adhering to the CNTsurface. These particles are believed to be amorphous carbon.30

After 30 min acid treatment, the CNT surface becomes cleanerand thinner, which proves that the short acid treatment odesimpact the purity of the structure. Furthermore, there is anenhancement of 50% for the BET surface area of the acid-treated specimen (Table 2). The pore size distribution ofspecimen is also modified during the treatment (see Figure 6b).Although the majority of the pores range from 2 to 4 nm for

both the as-prepared and acid-treated specimens, there is a widepeak in the range of 15−30 nm for the as-prepared specimen.After acid treatment, two smaller peaks around 7 and 10 nm areobserved. The work of Edwards et al.58 revealed that theconcentrated HNO3 treatment was not effective for purificationof the iron within the CNTs that serve as catalyst nanoparticles,although amorphous carbon can be eliminated. Thus, here, theincrease in the BET surface area and the change of pore sizedistribution is attributed to the etching of the carbon-baseddefects, such as amorphous carbon and very short, highlydefective CNTs, in the CNT fibers and films resulting insmaller pores and increasing specific surface area. These resultsfurther support the purification effect of acid treatment. Asdiscussed in the previous sections, amorphous carbon, shortCNTs and dangling ends, serving as resistances to thermalconduction and may prevent the overall thermal conductivityfrom achieving the expected performance. More effectivecontacts between nanotubes and less defective structure afterpurification greatly enhance the electrical and thermalconductivities.Additionally, the densities of the acid-treated CNT fibers and

films are larger than the as-prepared ones (Table 2), indicatingdensification of the macroscale structure as another effect ofacid treatment. Figure 7 shows the SEM images of the fracturemorphologies of the as-prepared and acid-treated CNT films.After acid treatment, the CNTs assembled into thicker bundleswith more compacted structure with larger densities. Since thecontact resistance depends strongly on the contact angle and

Table 2. Effects of Acid Treatment on Physical Properties ofCNT Fiber and Films

as-prepared acid-treated

sample fiber film fiber film

density (g cm−3) 0.67 1.19 0.71 1.60electrical conductivity(S cm−1)

700 1988 1622 4666

specific electricalconductivity (S cm2 g−1)

1045 1670 2285 2916

thermal conductivity(W m−1 K−1)

25 113 74 759

specific thermal conductivity(mW m2 kg−1 K−1)

37 104 94 474

BET surface area (m2 g−1) 78 124

Figure 6. Effects of acid treatment on (a) Raman spectra and (b) poresize distribution. TEM images of CNT (c) before and (d) after acidtreatment. Note the amorphous carbon at the surface of the as-synthesized CNT in part c, which is not observed in the acid-treatedsample in part d.

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.6b04114ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

G

intertube space,6 the considerable enhancement of conductiv-ities also can be attributed to the improving CNT−CNTcontact conductance due to the better alignment, shorterintertube space, and larger effective contact area afterdensification.Interfacial contact resistance can also be modified through

the covalently bonded nanotubes,59 and there may be someslightly functionalized CNTs existing after the short time acidtreatment.20,30 However, thorough functionalization wasreported to suppress the electrical conductivity of the CNTfibers20 and also decrease the intrinsic nanotube thermalconductivity60 due to the destruction of the crystalline structureof CNT, which can be observed by TEM. Based on ourcharacterizations of the structural modifications, functionaliza-tion is not the dominant reason for the observed enhancementof conductivities.As shown in Figure 8, the thermal conductivity and specific

thermal conductivity of our CNT fibers and films arecomparable to those of other CNT fibers and selected highlyconductive films.8,21−25,52,61−63 In particular, our acid-treatedCNT film combines both high absolute thermal conductivityand high specific thermal conductivity, which is even higherthan the typical values of conventional heat-transfer metals(copper, aluminum, and silver) and comparable to that of thehighest performing CNT fibers and films reported in theliterature. Interestingly, the specific thermal conductivity of ourCNT film is much higher than that of our CNT fiber, whichdiffers from the observation for the CNT fibers where thethermal conductivity increases nearly linearly with the

increasing density and the specific thermal conductivity remainsnearly constant in the density range of 0.11−0.87 g cm−3. Thisindicates some nonlinearity in the thermal conductivity withincreased density in a wider density range (e.g., the density ofthe as-prepared film is around 1.2 g cm−3). The nonlinearity inthe thermal conductivity with increased density was alsoreported by Marconnet et al.64 in their study of the thermalconductivity of aligned CNT−polymer nanocomposites as afunction of varying CNT volume fraction. The lower specificconductivity may be ascribed to the structure of the CNT fiberand the inter-CNT contact resistance. Since the CNT fiber wasrolling from the CNT film, it can be regarded as a cylinderstructure consisting of multiple layers of films. Aliev et al.21

studied the effect of multilayered structure on the thermalconductivity of CNT sheets stacked on top of one another.They found that increasing the number of layers decreased themeasured value of thermal conductivity, likely due to poorinterfacial transport between the sheets. The same effect ofdecreasing thermal conductivity with increasing layers occursfor graphene layers when they are stacked in graphite. In thatcase, interlayer interactions quench the thermal conductivity bynearly an order of magnitude. The lower conductivity ofmultilayered structure indicates the presence of enhancedscattering of heat carriers at the tube−tube interconnectionsbetween layers. This result suggests that an improvedfabrication method is required in the future to obtain highlydense CNT fibers with better intertube contacts and highthermal conductivity. In addition, the thermal conductivities inthis work are measured near room temperature. For manyapplications,65,66 it is of great importance to study the thermalconductivity of the CNT fiber across a range of temperatures.The technique developed here can be extended to enabletemperature-dependent measurements through control of thebase temperature of the fiber and the environment.

4. CONCLUSIONS

We demonstrate a robust, fast electrothermal technique forsimultaneously measuring the thermal conductivity of andconvection coefficient from CNT fibers and films. Themeasured thermal and electrical performance demonstratesthe promise for using these fibers and films in macroscaleapplications requiring effective heat dissipation such aselectrical cables and multifunctional composites.The thermal conductivity of the as-prepared fiber ranges

from 4.7 ± 0.3 to 28.0 ± 2.4 W m−1 K−1, which is much lowerthan the expected from the high thermal conductivity ofindividual CNTs, is comparable to other CNT-based structuresand rivals the specific thermal conductivity of metals. The

Figure 7. SEM images of as-prepared (a and b) and acid-treated (c andd) CNT film at different magnifications. Note that the acid-treatedspecimen has more compacted structure with thicker CNT bundlescompared with the more porous structure consisting thinner bundlesof the as-prepared samples. The denser structure contributes moreeffective intertube contacts to the enhanced thermal conductivity.

Figure 8. Thermal conductivity and specific thermal conductivity of this work compared with literature data for other CNT fibers and films made bydifferent processes.8,21−25,52,61−63

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.6b04114ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

H

major factors limiting the apparent thermal conductivity of theCNT fibers are the interfacial contact resistance and structuralinhomogeneity in the fibers.The convective heat transfer coefficient for the as-prepared

CNT fiber strongly depends on the diameter and volumefraction. A correlation between CNT volume fraction, diameter,and convective coefficient has been developed based on theexperimental data. New models considering the microstructureof CNT fiber are required in the future for better understandingthe mechanism of heat transfer in CNT based assemblies.Nitric acid treatment is demonstrated to be a simple, effective

approach to enhance the thermal conductivity of the CNTfibers and films. The thermal conductivity and specific thermalconductivity are enhanced by 670% and 450% respectively afteracid treatment. This great improvement is attributed to thereducing CNT−CNT contact resistance, modified fiberstructure, and increased density after the acid treatment. Theacid-tread CNT film demonstrates thermal conductivity of 759W m−1 K−1 and specific thermal conductivity of 474 mW m2

kg−1 K−1, which is even higher than the typical values ofconventional metals and comparable with the highly conductiveCNT assemblies reported in the literature. Compared to theCNT films, CNT fibers exhibit lower thermal conductivity andrequire the development of optimized fabrication method formore compacted fiber structures with improved intertubecontacts to further improve thermal performance.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acsami.6b04114.

Validation of one-dimensional heat transfer; method ofemissivity calibration and thermal image acquisition;influence of variable local heat transfer coefficient on thetemperature profile; influence of temperature onChurchill−Chu correlation; summary of the propertiesof the CNT fibers (PDF)

■ AUTHOR INFORMATIONCorresponding Authors*E-mail: mpedhm@nus.edu.sg (H.M.D.).*E-mail: amarconn@purdue.edu (A.M.M.).NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThe authors would like to thank Temasek Laboratory@NUSSingapore (R-394-001-077-232) for the financial support forthis project.

■ REFERENCES(1) Yu, M.-F.; Lourie, O.; Dyer, M. J.; Moloni, K.; Kelly, T. F.; Ruoff,R. S. Strength and Breaking Mechanism of Multiwalled CarbonNanotubes under Tensile Load. Science 2000, 287, 637−640.(2) Collins, P. G.; Avouris, P. Nanotubes for Electronics. Sci. Am.2000, 283, 62−69.(3) Pop, E.; Mann, D.; Wang, Q.; Goodson, K.; Dai, H. ThermalConductance of an Individual Single-Wall Carbon Nanotube aboveRoom Temperature. Nano Lett. 2006, 6, 96−100.(4) Lu, W.; Zu, M.; Byun, J. H.; Kim, B. S.; Chou, T. W. State of theArt of Carbon Nanotube Fibers: Opportunities and Challenges. Adv.Mater. 2012, 24, 1805−1833.

(5) Lekawa-Raus, A.; Patmore, J.; Kurzepa, L.; Bulmer, J.; Koziol, K.Electrical Properties of Carbon Nanotube Based Fibers and TheirFuture Use in Electrical Wiring. Adv. Funct. Mater. 2014, 24, 3661−3682.(6) Marconnet, A. M.; Panzer, M. A.; Goodson, K. E. ThermalConduction Phenomena in Carbon Nanotubes and Related Nano-structured Materials. Rev. Mod. Phys. 2013, 85, 1295.(7) Vigolo, B.; Penicaud, A.; Coulon, C.; Sauder, C.; Pailler, R.;Journet, C.; Bernier, P.; Poulin, P. Macroscopic Fibers and Ribbons ofOriented Carbon Nanotubes. Science 2000, 290, 1331−1334.(8) Behabtu, N.; Young, C. C.; Tsentalovich, D. E.; Kleinerman, O.;Wang, X.; Ma, A. W.; Bengio, E. A.; ter Waarbeek, R. F.; de Jong, J. J.;Hoogerwerf, R. E.; et al. Strong, Light, Multifunctional Fibers ofCarbon Nanotubes with Ultrahigh Conductivity. Science 2013, 339,182−186.(9) Zhang, M.; Atkinson, K. R.; Baughman, R. H. MultifunctionalCarbon Nanotube Yarns by Downsizing an Ancient Technology.Science 2004, 306, 1358−1361.(10) Li, Y.-L.; Kinloch, I. A.; Windle, A. H. Direct Spinning ofCarbon Nanotube Fibers from Chemical Vapor Deposition Synthesis.Science 2004, 304, 276−278.(11) Vilatela, J. J.; Windle, A. H. Yarn - Like Carbon NanotubeFibers. Adv. Mater. 2010, 22, 4959−4963.(12) Liu, P.; Lam, A.; Fan, Z.; Tran, T. Q.; Duong, H. M. AdvancedMultifunctional Properties of Aligned Carbon Nanotube-Epoxy ThinFilm Composites. Mater. Des. 2015, 87, 600−605.(13) Liu, P.; Tran, T. Q.; Fan, Z.; Duong, H. M. FormationMechanisms and Morphological Effects on Multi-Properties of CarbonNanotube Fibers and Their Polyimide Aerogel-Coated Composites.Compos. Sci. Technol. 2015, 117, 114−120.(14) Tran, T. Q.; Fan, Z.; Liu, P.; Myint, S. M.; Duong, H. M. Super-Strong and Highly Conductive Carbon Nanotube Ribbons from Post-Treatment Methods. Carbon 2016, 99, 407−415.(15) Koziol, K.; Vilatela, J.; Moisala, A.; Motta, M.; Cunniff, P.;Sennett, M.; Windle, A. High-Performance Carbon Nanotube Fiber.Science 2007, 318, 1892−1895.(16) Wang, J.; Luo, X.; Wu, T.; Chen, Y. High-Strength CarbonNanotube Fibre-Like Ribbon with High Ductility and High ElectricalConductivity. Nat. Commun. 2014, DOI: 10.1038/ncomms4848.(17) Miao, M. Electrical Conductivity of Pure Carbon NanotubeYarns. Carbon 2011, 49, 3755−3761.(18) Gspann, T.; Smail, F.; Windle, A. Spinning of Carbon NanotubeFibres Using the Floating Catalyst High Temperature Route: PurityIssues and the Critical Role of Sulphur. Faraday Discuss. 2014, 173,47−65.(19) Zhao, Y.; Wei, J.; Vajtai, R.; Ajayan, P. M.; Barrera, E. V. IodineDoped Carbon Nanotube Cables Exceeding Specific ElectricalConductivity of Metals. Sci. Rep. 2011, DOI: 10.1038/srep00083.(20) Meng, F.; Zhao, J.; Ye, Y.; Zhang, X.; Li, Q. Carbon NanotubeFibers for Electrochemical Applications: Effect of Enhanced Interfacesby an Acid Treatment. Nanoscale 2012, 4, 7464−7468.(21) Aliev, A. E.; Guthy, C.; Zhang, M.; Fang, S.; Zakhidov, A. A.;Fischer, J. E.; Baughman, R. H. Thermal Transport in MWCNTSheets and Yarns. Carbon 2007, 45, 2880−2888.(22) Jakubinek, M. B.; Johnson, M. B.; White, M. A.; Jayasinghe, C.;Li, G.; Cho, W.; Schulz, M. J.; Shanov, V. Thermal and ElectricalConductivity of Array-Spun Multi-Walled Carbon Nanotube Yarns.Carbon 2012, 50, 244−248.(23) Mayhew, E.; Prakash, V. Thermal Conductivity of HighPerformance Carbon Nanotube Yarn-Like Fibers. J. Appl. Phys.2014, 115, 174306.(24) Ericson, L. M.; Fan, H.; Peng, H.; Davis, V. A.; Zhou, W.;Sulpizio, J.; Wang, Y.; Booker, R.; Vavro, J.; Guthy, C. Macroscopic,Neat, Single-Walled Carbon Nanotube Fibers. Science 2004, 305,1447−1450.(25) Imaizumi, S.; Matsumoto, H.; Konosu, Y.; Tsuboi, K.;Minagawa, M.; Tanioka, A.; Koziol, K.; Windle, A. Top-Down ProcessBased on Electrospinning, Twisting, and Heating for Producing One-

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.6b04114ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

I

Dimensional Carbon Nanotube Assembly. ACS Appl. Mater. Interfaces2011, 3, 469−475.(26) Zhou, W.; Vavro, J.; Guthy, C.; Winey, K. I.; Fischer, J. E.;Ericson, L. M.; Ramesh, S.; Saini, R.; Davis, V. A.; Kittrell, C. SingleWall Carbon Nanotube Fibers Extruded from Super-Acid Suspensions:Preferred Orientation, Electrical, and Thermal Transport. J. Appl. Phys.2004, 95, 649−655.(27) Badaire, S.; Pichot, V.; Zakri, C.; Poulin, P.; Launois, P.; Vavro,J.; Guthy, C.; Chen, M.; Fischer, J. E. Correlation of Properties withPreferred Orientation in Coagulated and Stretch-Aligned Single-WallCarbon Nanotubes. J. Appl. Phys. 2004, 96, 7509−7513.(28) James, M.; Windle, A. H. Measuring the Axial and RadialThermal Conductivities of Carbon Nanotube Fibers. Presented atNT12 Thirteenth International conference on Science and Applicationof Nanotubes, Brisbane, Australia, June, 2012.(29) Bergman, T. L.; Incropera, F. P.; Lavine, A. S. Fundamentals ofHeat and Mass Transfer, Seventh ed.; John Wiley & Sons, 2011.(30) Liu, C.; Fan, S. Effects of Chemical Modifications on theThermal Conductivity of Carbon Nanotube Composites. Appl. Phys.Lett. 2005, 86, 123106.(31) Li, Q.; Li, Y.; Zhang, X.; Chikkannanavar, S. B.; Zhao, Y.;Dangelewicz, A. M.; Zheng, L.; Doorn, S. K.; Jia, Q.; Peterson, D. E.;et al. Structure - Dependent Electrical Properties of Carbon NanotubeFibers. Adv. Mater. 2007, 19, 3358−3363.(32) Kodama, T.; Park, W.; Marconnet, A.; Lee, J.; Asheghi, M.;Goodson, K. E. In-Plane Thermal Conductivity Measurement onNanoscale Conductive Materials with on-Substrate Device Config-uration. 13th IEEE Intersoc. Conf. Therm. Thermomechan. Phenom.Electron. Syst. (ITherm) 2012, 250−255.(33) Pop, E.; Mann, D. A.; Goodson, K. E.; Dai, H. Electrical andThermal Transport in Metallic Single-Wall Carbon Nanotubes onInsulating Substrates. J. Appl. Phys. 2007, 101, 093710.(34) Alvarenga, J.; Jarosz, P. R.; Schauerman, C. M.; Moses, B. T.;Landi, B. J.; Cress, C. D.; Raffaelle, R. P. High Conductivity CarbonNanotube Wires from Radial Densification and Ionic Doping. Appl.Phys. Lett. 2010, 97, 182106.(35) Liu, K.; Sun, Y.; Zhou, R.; Zhu, H.; Wang, J.; Liu, L.; Fan, S.;Jiang, K. Carbon Nanotube Yarns with High Tensile Strength Made bya Twisting and Shrinking Method. Nanotechnology 2010, 21, 045708.(36) Balandin, A. A. Thermal Properties of Graphene andNanostructured Carbon Materials. Nat. Mater. 2011, 10, 569−581.(37) Chalopin, Y.; Volz, S.; Mingo, N. Upper Bound to the ThermalConductivity of Carbon Nanotube Pellets. J. Appl. Phys. 2009, 105,084301.(38) Pettes, M. T.; Shi, L. Thermal and Structural Characterizationsof Individual Single-, Double-, and Multi-Walled Carbon Nanotubes.Adv. Funct. Mater. 2009, 19, 3918−3925.(39) Choi, T. Y.; Poulikakos, D.; Tharian, J.; Sennhauser, U.Measurement of Thermal Conductivity of Individual MultiwalledCarbon Nanotubes by the 3-W Method. Appl. Phys. Lett. 2005, 87,13108−13108.(40) Choi, T.-Y.; Poulikakos, D.; Tharian, J.; Sennhauser, U.Measurement of the Thermal Conductivity of Individual CarbonNanotubes by the Four-Point Three-Ω Method. Nano Lett. 2006, 6,1589−1593.(41) Fujii, M.; Zhang, X.; Xie, H.; Ago, H.; Takahashi, K.; Ikuta, T.;Abe, H.; Shimizu, T. Measuring the Thermal Conductivity of a SingleCarbon Nanotube. Phys. Rev. Lett. 2005, 95, 065502.(42) Aliev, A. E.; Lima, M. H.; Silverman, E. M.; Baughman, R. H.Thermal Conductivity of Multi-Walled Carbon Nanotube Sheets:Radiation Losses and Quenching of Phonon Modes. Nanotechnology2010, 21, 035709.(43) Li, Q.; Liu, C.; Wang, X.; Fan, S. Measuring the ThermalConductivity of Individual Carbon Nanotubes by the Raman ShiftMethod. Nanotechnology 2009, 20, 145702.(44) Kim, P.; Shi, L.; Majumdar, A.; McEuen, P. Thermal TransportMeasurements of Individual Multiwalled Nanotubes. Phys. Rev. Lett.2001, 87, 215502.

(45) Ruoff, R. S.; Tersoff, J.; Lorents, D. C.; Subramoney, S.; Chan,B. Radial Deformation of Carbon Nanotubes by Van Der WaalsForces. Nature 1993, 364, 514−516.(46) Evans, W. J.; Shen, M.; Keblinski, P. Inter-Tube ThermalConductance in Carbon Nanotubes Arrays and Bundles: Effects ofContact Area and Pressure. Appl. Phys. Lett. 2012, 100, 261908.(47) Prasher, R. S.; Hu, X.; Chalopin, Y.; Mingo, N.; Lofgreen, K.;Volz, S.; Cleri, F.; Keblinski, P. Turning Carbon Nanotubes fromExceptional Heat Conductors into Insulators. Phys. Rev. Lett. 2009,102, 105901.(48) Volkov, A. N.; Zhigilei, L. V. Heat Conduction in CarbonNanotube Materials: Strong Effect of Intrinsic Thermal Conductivityof Carbon Nanotubes. Appl. Phys. Lett. 2012, 101, 043113.(49) Salvetat, J.-P.; Kulik, A. J.; Bonard, J.-M.; Briggs, G. A. D.;Stockli, T.; Metenier, K.; Bonnamy, S.; Beguin, F.; Burnham, N. A.;Forro, L. Elastic Modulus of Ordered and Disordered MultiwalledCarbon Nanotubes. Adv. Mater. 1999, 11, 161−165.(50) Peirs, J.; Reynaerts, D.; Van Brussel, H. In Scale Effects andThermal Considerations for Micro-Actuators. Proceedings of 1998 IEEEInternational Conference on Robotics and Automation, Leuven, Belgium;IEEE: 1998; pp 1516−1521.(51) Alam, M.; Raghu, A.; Haque, M.; Muratore, C.; Voevodin, A.Structural Size and Temperature Dependence of Solid to Air HeatTransfer. Int. J. Therm. Sci. 2013, 73, 1−7.(52) Li, M.; Li, C.; Wang, J.; Xiao, X.; Yue, Y. Parallel Measurementof Conductive and Convective Thermal Transport of Micro/Nanowires Based on Raman Mapping. Appl. Phys. Lett. 2015, 106,253108.(53) Hsu, I.-K.; Pettes, M. T.; Aykol, M.; Chang, C.-C.; Hung, W.-H.;Theiss, J.; Shi, L.; Cronin, S. B. Direct Observation of Heat Dissipationin Individual Suspended Carbon Nanotubes Using a Two-LaserTechnique. J. Appl. Phys. 2011, 110, 044328.(54) Kim, K. J.; King, W. P. Thermal Conduction between a HeatedMicrocantilever and a Surrounding Air Environment. Appl. Therm.Eng. 2009, 29, 1631−1641.(55) Wang, Z.; Tang, D. Investigation of Heat Transfer aroundMicrowire in Air Environment Using 3ω Method. Int. J. Therm. Sci.2013, 64, 145−151.(56) Guan, N.; Liu, Z.; Zhang, C.; Jiang, G. Natural Convection HeatTransfer on Surfaces of Copper Micro-Wires. Heat Mass Transfer2014, 50, 275−284.(57) Hu, X. J.; Jain, A.; Goodson, K. E. Investigation of the NaturalConvection Boundary Condition in Microfabricated Structures. Int. J.Therm. Sci. 2008, 47, 820−824.(58) Edwards, E.; Antunes, E.; Botelho, E. C.; Baldan, M.; Corat, E.Evaluation of Residual Iron in Carbon Nanotubes Purified by AcidTreatments. Appl. Surf. Sci. 2011, 258, 641−648.(59) Varshney, V.; Patnaik, S. S.; Roy, A. K.; Farmer, B. L. Modelingof Thermal Conductance at Transverse CNT-CNT Interfaces. J. Phys.Chem. C 2010, 114, 16223−16228.(60) Shenogin, S.; Bodapati, A.; Xue, L.; Ozisik, R.; Keblinski, P.Effect of Chemical Functionalization on Thermal Transport of CarbonNanotube Composites. Appl. Phys. Lett. 2004, 85, 2229−2231.(61) Zhang, L.; Zhang, G.; Liu, C.; Fan, S. High-Density CarbonNanotube Buckypapers with Superior Transport and MechanicalProperties. Nano Lett. 2012, 12, 4848−4852.(62) Gonnet, P.; Liang, Z.; Choi, E. S.; Kadambala, R. S.; Zhang, C.;Brooks, J. S.; Wang, B.; Kramer, L. Thermal Conductivity ofMagnetically Aligned Carbon Nanotube Buckypapers and Nano-composites. Curr. Appl. Phys. 2006, 6, 119−122.(63) Tong, T.; Zhao, Y.; Delzeit, L.; Kashani, A.; Meyyappan, M.;Majumdar, A. Dense Vertically Aligned Multiwalled Carbon NanotubeArrays as Thermal Interface Materials. IEEE Trans. Compon. Packag.Technol. 2007, 30, 92.(64) Marconnet, A. M.; Yamamoto, N.; Panzer, M. A.; Wardle, B. L.;Goodson, K. E. Thermal Conduction in Aligned Carbon Nanotube-Polymer Nanocomposites with High Packing Density. ACS Nano2011, 5, 4818−4825.

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.6b04114ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

J

(65) Fairchild, S. B.; Bulmer, J. S.; Sparkes, M.; Boeckl, J.; Cahay, M.;Back, T.; Murray, P. T.; Gruen, G.; Lange, M.; Lockwood, N. P.; et al.Field Emission from Laser Cut CNT Fibers and Films. J. Mater. Res.2014, 29, 392−402.(66) Janas, D.; Cabrero-Vilatela, A.; Bulmer, J.; Kurzepa, L.; Koziol,K. K. Carbon Nanotube Wires for High-Temperature Performance.Carbon 2013, 64, 305−314.

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.6b04114ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

K