PHYSICS APPLIED LETTERS 89, 133102 2006
Effect of single-walled carbon nanotube purity on the thermal conductivity of carbon nanotube-based composites
Aiping Yu, Mikhail E. Itkis, Elena Bekyarova, and Robert C. Haddona
Center for Nanoscale Science and Engineering, University of California, Riverside, California 92521-0403; Department of Chemistry, University of California, Riverside, California 92521-0403; and Department of Chemical and Environmental Engineering, University of California, Riverside, California 92521-0403
Received 6 July 2006; accepted 4 August 2006; published online 25 September 2006 Raw as-prepared AP and purified single-walled carbon nanotubes SWNTs were utilized for the preparation of SWNT-epoxy composites. Purified functionalized SWNTs provide a significantly greater enhancement of the thermal conductivity, whereas AP-SWNTs allow the best electrical properties because of their ability to form efficient percolating network. A series of SWNT samples of varying purity but identical chemical functionality was prepared to delineate the effect of SWNT purity on the thermal conductivity of SWNT-epoxy composites. The authors found that purified SWNTs provide approximately five times greater enhancement of the thermal conductivity than the impure SWNT fraction demonstrating the significance of SWNT quality for thermal management. 2006 American Institute of Physics. DOI: 10.1063/1.2357580 The outstanding thermal and mechanical properties of carbon nanotubes CNTs offer the possibility of a new generation of thermal interface materials TIMs for thermal management of high density electronics. The thermal conductivity of individual CNTs up to 3000 W / mK 1 is significantly higher than that of traditionally utilized fillers, such as alumina, silver, copper, carbon black, and carbon fibers. Moreover, CNTs provide a more efficient network for heat flow inside the polymer matrix because of their high aspect ratio.2 Several groups have studied the thermal properties of single-walled carbon nanotube SWNTs -based composite materials38 and some of the publications report an enhancement of the thermal conductivity of over 100% per 1 wt % of CNT loading,3,4,7 although in some cases no enhancement was observed.6 Intrinsic to the performance of SWNT-based composites are the quality of dispersion of SWNTs in the polymer matrix, SWNT loading, conformation and alignment, and the thermal resistance of the SWNT/matrix interface.311 While these intrinsic variables are extremely important, we found that an extrinsic factor--the SWNT purity--remains critical in defining the thermal performance of SWNT/epoxy composites. In this letter we present a systematic study of the effect of the purity of the SWNT material on the thermal performance of SWNT-based composites. In this study as-prepared arc discharge produced SWNTs AP-SWNTs from Carbon Solutions, Inc. were treated with nitric acid to give purified SWNTs P-SWNTs .12 For composite preparation the SWNTs were dispersed ultrasonically for 5 h in acetone, and the epoxy resin, diglycidyl ether of bisphenol A EPON 862 , was subsequently added to the SWNT suspension followed by 30 min high-shear mixing to ensure good dispersion. The residual solvent was removed at 50 C in a vacuum oven and an aromatic diamine curing agent diethyltoluenediamine, EPI-cure W was added under continuous stirring with a homogenizer. The mixture was loaded in a custom-made stainless steel mold, degassed and heated in vacuum at 100 C for 2 h and 150 C for another
a
2 h to complete the curing cycle. The thermal conductivity of disk shaped CNT/epoxy composite samples of 1 in. diameter Fig. 1, inset was measured with a FOX50 LaserComp, Inc steady-state heat flow measurement apparatus, employing a two thickness measurement cycle in order to eliminate thermal contact resistance to the sample. Figure 1 shows the thermal conductivity enhancement, / 0 % , of SWNT/epoxy composites prepared with APrelative purity RP = 60% and P-SWNT RP= 150% as a function of the nanotube loading, where 0 is the thermal is conductivity of the neat polymer 0.201 W / mK and the change of the thermal conductivity of the composite due to addition of SWNTs. The enhancement observed for P-SWNT-based composite is 80% higher than that for the analogous composite prepared with AP-SWNTs. Moreover, P-SWNTs are much more dispersible in the epoxy resin, while AP-SWNTs produce a significantly stronger increase in viscosity than P-SWNTs when added to the epoxy. As a result, it is very difficult to achieve an AP-SWNT loading higher than 5 wt %, while for P-SWNTs loadings higher than 10 wt % are readily achievable. The superior performance of P-SWNTs in comparison with the AP-SWNT material may be attributed to a combination of two factors. First, the carboxylic acid groups introduced to the SWNTs during the nitric acid treatment materi-
Author to whom correspondence should be addressed; electronic mail: [e-mail]
FIG. 1. Color online Thermal conductivity enhancement, / 0 % , of SWNT/epoxy composites as a function of AP- and P-SWNT loading. The inset shows 1 in. diameter disk samples of neat epoxy and SWNT/epoxy composites. 2006 American Institute of Physics
0003-6951/2006/89 13 /133102/3/$23.00
89, 133102-1
133102-2
Yu et al.
Appl. Phys. Lett. 89, 133102 2006
FIG. 2. Color online a d Characterization of the purity of four SWNT fractions utilized for SWNT/epoxy composites by NIR spectroscopy. The SWNT relative purity RP is proportional to the ratio of the area of the spectral feature corresponding to the second interband transition of semiconducting SWNTs, after a base line correction a d , top panels , to the total area under the spectral curve bottom panels ; e h AFM images of the same four SWNT fractions; i l SEM images of SWNT/epoxy composites prepared using the same four SWNT fractions; and m thermal conductivity enhancement for the SWNT/epoxy composites prepared using the same four SWNT fractions and 4 wt % carbon material loading.
FIG. 3. Color online a Thermal conductivity of SWNT/epoxy composites as a function of SWNT volume fraction of AP- and P-SWNTs. Theoretical values are shown as dashed straight lines: 1 Ref. 11 and 2 Ref. 2 . b Electrical conductivities of SWNT/epoxy composites as a function of for AP- and P-SWNTs. The curves correspond to fits obtained using the - c t with best fitting parameters: threshpercolating network model old volume fraction c = 0.04% 1.0% and critical exponent t = 1.7 2.4 for AP-SWNTs P-SWNTs . The inset shows the same fit in log vs log - c plots; AFM images of c AP-SWNTs; and d P-SWNTs at similar SWNT concentrations showing the tendency of AP-SWNTs to form extended clusters more readily connected in the percolating network than the well-dispersed P-SWNTs.
ally aid the dispersal of SWNTs in many matrices and may be expected to cross-link with the polymer matrix13 and this is expected to reduce the phonon scattering at the SWNT/ polymer interface.10 Second, the improved thermal conductivity observed in the composites fabricated with P-SWNTs may originate simply from the higher purity of the material, which results in higher effective concentration of SWNTs. In order to examine the role played by these two factors we prepared a series of SWNT/epoxy composites based on SWNT samples with common chemical functionalization but a different level of purity. The purity of the carbonaceous component of the SWNT sample was measured with reference to an arbitrary standard using the solution phase near infrared NIR spectroscopy Figs. 2 a 2 d to give a RP.14,15 SWNT samples of varying purities were obtained from different stages of the bulk scale purification procedure of SWNTs.12,16 Briefly, the first step of the procedure includes the reflux of AP-SWNTs RP= 60% in nitric acid. This step exfoliates the SWNT bundles and removes a significant amount of the metal impurities; according to the NIR analysis Fig. 2 b the overall carbonaceous purity remains at about the same level RP 60% because both SWNTs and amorphous carbon are attacked during the oxidation process. The second step is a low-speed which centrifugation removes most of the amorphous carbon and the RP increases to
100% Fig. 2 c . The next purification step relies on a highspeed centrifugation, which produces two fractions: a supernatant containing mostly SWNTs with RP= 150% Fig. 2 d and a precipitate containing mostly graphitic nanoparticles with a trace amount of SWNTs RP 10% Fig. 2 a .16 The atomic force microscopy AFM Figs. 2 e 2 h and scanning electron microscopy SEM Figs. 2 i 2 l images clearly confirm the NIR findings. Thus the process yields four fractions of SWNT material of similar functionality and with RP varying from 10% to 150%, and these samples have been used as fillers to prepare four sets of SWNT/epoxy composite samples at the same loading of 4 wt % carbon material. The thermal conductivity enhancement TCE for the four composite samples is presented in Fig. 2 m . As the SWNT purity changes from 10% to 150% the TCE is increased from 45% to 200%. To rationalize this result we should note that the SWNT fractions utilized in this experiment as well as all commercially available SWNT samples contain, in addition to SWNTs, a large quantity of impurities in the form of amorphous carbon and graphitic nanoparticles.14,15 As expected, SWNTs are a much more efficient filler for heat conduction in composites because of their significantly higher intrinsic thermal conductivity and higher aspect ratio compared to carbonaceous nanoparticles.15,711 Figures 3 a and 3 b compare the behavior of the thermal and electrical conductivities of the SWNT/epoxy composites with increasing the SWNT loading for AP- and P-SWNTs. The electrical conductivity Fig. 3 b shows an increase of several orders of magnitude as SWNTs are introduced into the composite and follows the behavior expected for the formation of a percolating network:6,17,18 - c t, where and c describe the volume fraction
133102-3
Yu et al.
Appl. Phys. Lett. 89, 133102 2006
of conducting filler and its critical percolating value, respectively, and t is the critical exponent. For AP-SWNTs the percolating threshold c = 0.04% was found to be much lower than for P-SWNTs, c = 1%. The electrical conductivity of AP-SWNT-based composites is several orders of magnitude higher than that for P-SWNTs composites at the same nanotube loading Fig. 3 b , contrary to the case of thermal conductivity Fig. 3 a . In contrast with the electrical conductivity behavior, we did not observe a percolation threshold in the thermal conductivity18 and this highlights the differences in the phenomena under investigation. In the case of the thermal transport, the conductivity increases only by a factor of 5 over the full range of SWNT loadings 0%8% Fig. 3 a , whereas in the absence of the carbon nanotubes the epoxy is effectively an electrical insulator and the electrical conductivity increases by ten orders of magnitude over the same range of SWNT loading Fig. 3 b . This is in accord with the significantly larger difference in the electrical conductivities of the components in comparison with their thermal conductivities: -14 S / cm and SWNT 103 S / cm,16,19,20 polymer polymer 10 = 0.2 W / mK, and SWNT = 10 100 W / mK.19 In addition, the large thermal interface resistance at the SWNT junctions corresponding to the involvement of the matrix to the heat flow acts to suppress the formation of a percolating network in the case of the thermal transport.9 It is interesting to note the difference in the relative values of the thermal and electrical transport properties in AP and purified SWNTs Figs. 3 a and 3 b , which we attribute to the nature of the dispersions and the transport processes involved. As can be seen in the AFM images in Figs. 3 c and 3 d AP SWNTs have a tendency to form extended clusters of large SWNT bundles Fig. 3 c , which are more readily connected in the percolating network17 resulting in a low percolation threshold, perhaps as a result of phase separation due to their poor solvation by the matrix. The purified SWNTs are well exfoliated into small bundles, which are individually dispersed in the matrix Figs. 2 k and 2 l due to the compatibility of the functional groups and polymer matrix thus impeding the formation of percolating chains.18 The data presented in this letter clearly demonstrate the dramatic effect of the purity of SWNTs on the thermal performance of SWNT-based composites, although even for highly purified SWNT samples the results obtained in this and previous reports38 are below the theoretically predicted thermal conductivities Fig. 3 a .2,11 The largest deviation from the theoretical predictions is observed at high SWNT loading where the dependence of thermal conductivity on SWNT loading is sublinear presumably due to bending of the CNTs inside the matrix, which reduces the favorable aspect ratio8 and phase separation. These problems and acoustic phonon mismatch can be addressed by introducing a better developed interfacial bonding between the CNTs and the polymer matrix, which can be engineered by carefully
matching the functional groups in the CNTs with the chemical structure of the polymer10,13 by utilizing the chemistry of SWNTs.13,21 In conclusion, we demonstrated that the purity of the starting SWNT material dramatically affects the thermal conductivity enhancement of SWNT/epoxy composites. Graphitic and carbonaceous impurities are low efficiency fillers, which lead to an increased viscosity of the SWNT/polymer dispersion without improving the thermal conductivity of the SWNT/epoxy composites. Even for highly purified SWNT material the thermal conductivity of the composites is below the theoretical values and further improvements of the SWNT-polymer matrix interface and the SWNT aspect ratio are required in order to develop efficient CNT-based TIMs for high density electronic packaging. This research was supported by the Defense Microelectronics Activity DMEA under Agreement No. 23109CNN05--DMEA H94003-05-2-0505.
P. Kim, L. Shi, A. Majumdar, and P. L. McEuen, Phys. Rev. Lett. 87, 215502 2001 . 2 L. E. Nielsen, Ind. Eng. Chem. Fundam. 13, 17 1974 . 3 M. J. Biercuk, M. C. Llaguno, M. Radosavljevic, J. K. Hyun, A. T. Johnson, and J. E. Fischer, Appl. Phys. Lett. 80, 2767 2002 . 4 E. S. Choi, J. S. Brooks, D. L. Eaton, M. S. Al-Haik, M. Y. Hussaini, H. Garmestani, D. Li, and K. Dahmen, J. Appl. Phys. 94, 6034 2003 . 5 C. H. Liu, H. Huang, Y. Wu, and S. S. Fan, Appl. Phys. Lett. 84, 4248 2004 . 6 A. Moisala, Q. Li, I. A. Kinloch, and A. H. Windle, Compos. Sci. Technol. 66, 1285 2005 . 7 H. Huang, C. H. Liu, Y. Wu, and S. S. Fan, Adv. Mater. Weinheim, Ger. 17, 1652 2005 . 8 P. C. Song, C. H. Liu, and S. S. Fan, Appl. Phys. Lett. 88, 153111 2006 . 9 S. Shenogin, L. P. Xue, R. Ozisik, P. Keblinski, and D. G. Cahill, J. Appl. Phys. 95, 8136 2004 . 10 S. Shenogin, A. Bodapati, L. Xue, R. Ozisik, and P. Keblinski, Appl. Phys. Lett. 85, 2229 2004 . 11 C. W. Nan, Z. Shi, and Y. Lin, Chem. Phys. Lett. 375, 666 2003 . 12 H. Hu, A. Yu, E. Kim, B. Zhao, M. E. Itkis, E. Bekyarova, and R. C. Haddon, J. Phys. Chem. B 109, 11520 2005 . 13 J. Gao, B. Zhao, M. E. Itkis, E. Bekyarova, H. Hu, V. Kranak, A. Yu, and R. C. Haddon, J. Am. Chem. Soc. 128, 7492 2006 . 14 R. C. Haddon, J. Sippel, A. G. Rinzler, and F. Papadimitrakopoulos, MRS Bull. 29, 252 2004 . 15 M. E. Itkis, D. Perea, R. Jung, S. Niyogi, and R. C. Haddon, J. Am. Chem. Soc. 127, 3439 2005 . 16 A. Yu, E. Bekyarova, M. E. Itkis, D. Fakhrutdinov, R. Webster, and R. C. Haddon, J. Am. Chem. Soc. 128, 9902 2006 . 17 M. B. Bryning, M. F. Islam, J. M. Kikkawa, and A. G. Yodh, Adv. Mater. Weinheim, Ger. 17, 1186 2005 . 18 M. B. Bryning, D. E. Milkie, M. F. Islam, J. M. Kikkawa, and A. G. Yodh, Appl. Phys. Lett. 87, 161909 2005 . 19 J. Hone, M. C. Llaguno, N. M. Nemes, A. T. Johnson, J. E. Fischer, D. A. Walters, M. J. Casavant, J. Schmidt, and R. E. Smalley, Appl. Phys. Lett. 77, 666 2000 . 20 E. Bekyarova, M. E. Itkis, N. Cabrera, B. Zhao, A. Yu, J. Gao, and R. C. Haddon, J. Am. Chem. Soc. 127, 5990 2005 . 21 S. Niyogi, M. A. Hamon, H. Hu, B. Zhao, P. Bhowmik, R. Sen, M. E. Itkis, and R. C. Haddon, Acc. Chem. Res. 35, 1105 2002 .
1