Enhanced thermal conductivity of epoxy/three-dimensional carbon hybrid filler composites for effective heat dissipation

Abstract

Graphitic carbon nanomaterials (CNMs) are recognized as next-generation heat dissipating materials (HDMs) for efficient thermal conduction within a polymer composite. Commercially used carbon-based HDMs, including carbon blacks, carbon nanotubes (CNTs), and graphites, are limited by low thermal conductivity under 50% filler content. Two-dimensional graphenes show high thermal conductivity in their major (xy) planes; however, they still exhibit low thermal conductivity in the z direction, i.e., perpendicular to the major plane, because of the difficulty in obtaining a proper vertical alignment. Here, we introduce a straightforward strategy to improve the thermal conductivity of graphene-based HDMs in both the xy- and z-directions and report the results of our investigation of the thermal conductivity behavior of the epoxy composites. Our newly designed graphene-based HDMs were observed to adopt directly anchored and ohmic-contact morphologies between graphene nanoplatelets (GNP) and CNTs via bimetallic nanoparticle decoration on the GNP surface and subsequent carbon vapor deposition (CVD), and their epoxy composites showed an approximately two-fold enhancement of both in- and through-plane thermal conductivities compared to those of bare GNPs.

---

1. Introduction

Thermal management techniques for effective heat dissipation of high-density electronic devices, which have become quite popular, are critical owing to the damaging effects of the heat that is generated when operating these devices and hence their internal components, which include batteries, central processing units (CPUs), radio frequency (RF) units, and displays.^1–3 The critical need for such techniques is highlighted by the observation that the lifetimes of electronic devices are shortened by up to 1500–2000 hours whenever the temperature in the device rises by 1 °C.⁴ Examples of heat-dissipating systems required in electronic devices include heat-dissipating printed circuit boards (PCBs), heat-spreading sheets, thermal interface materials (TIMs), and thermal radiation coating between the heat source and the heat sink.^5–8 To date, the various heat-dissipating materials (HDMs) have been divided into those that are electrically insulating (e.g., SiO₂, SiC, Al₂O₃, BN, AlN, MgO)^9–11 and those that are conducting (e.g., Cu, Al, Ni, Ag, graphite, carbon black, carbon fiber),^12–14 which have been generally employed as fillers in a light-weight polymer matrix to construct the heat-dissipating systems. However, to attain sufficient thermal conductivity for commercial applications (>5 W mK⁻¹), high contents (>50%) of these conventional HDMs have to be used as filler in the polymer matrix,^6,15,16 which results in the reduction of the mechanical robustness of the polymer composite.

Recently, graphitic carbon nanomaterials (CNMs) including carbon nanotubes (CNTs), graphenes, and multi-layered graphene nanoplatelets (GNPs) have been recognized as next-generation HDMs because of their excellent properties such as light weight, tremendous thermal conductivity (3000–6000 W mK⁻¹), mechanical robustness (Young's modulus ∼ 1 Ta), and good dispersibility.¹⁷ In particular, graphenes, which are very large in two dimensions, enable large areas of contact between neighbors in a polymer composite, resulting in the highest thermal conductivity among the various CNMs.^18–20 However, since two-dimensional graphitic layers are easily re-stacked by the influence of van der Waals forces, graphenes tend to be oriented horizontally on the substrate, such that anisotropic thermal conductivity behavior occurs between the directions of the xy-dimension (in-plane) and the z-axis (through-plane) relative to the substrate.^21,22 In particular, since vertically aligned graphenes within the polymer matrix are rarely prepared and difficult to form, development of newly designed CNMs via hybridization is important for enhancing through-plane as well as in-plane thermal conductivity. Indeed, the main direction of the heat transfer in the device is one-dimensional heat flow between the heat source and the surroundings.

In previous research, CNTs, carbon fibers, and metal nanowires, which have high thermal conductivities along the z-axis, were incorporated with graphenes, by simple mixing or direct synthesis, to elevate their thermal conducting performances.^23–28 Among these one-dimensional fillers, CNTs have been considered as an ideal supplementary material because of their high thermal conductivity (3000–6600 W mK⁻¹), comparable to that of graphenes.²⁹ However, in this hybridization, many of the CNTs were observed to form a point-type contact geometry^23,26,30 or extra buffer layers on the graphenes prohibiting direct carbon–carbon contacts,³⁰ which led to a decrease of thermal conducting performances such that enhancement of the thermal conductivity hardly occurred in both the in- and through-plane directions. In our previous work,^30,31 we investigated GNP-CNT hybrid fillers via CNT synthesis on the GNP surface employing an MgO buffer layer as a metal catalyst support; MgO-coated GNPs, however, reduced in-plane thermal conductivity owing to the high interfacial thermal resistance between MgO (45–60 W mK⁻¹) and GNP, even though through-plane thermal conductivity was enhanced by the vertically-aligned CNTs in an epoxy composite.

In this study, we demonstrate a straightforward strategy to combine GNPs with CNTs, without any buffer layers, to improve both in- and through-plane thermal conductivities in an epoxy composite compared with those of individual fillers. Electroless plating, which enables simple, controllable, and selective deposition of metals consisting of one more components regardless of the substrate,³² was used to decorate GNP surfaces with transition metal nanoparticles. CNTs were then successfully combined with the GNP surface using chemical vapor deposition (CVD). An epoxy composite with 20 wt% three-dimensional carbon hybrid fillers (3-D CHFs) showed 3.2 times, 2.1 times, and 1.7 times greater through-plane thermal conductivity and 4.3 times, 1.7 times, and 1.9 times greater in-plane thermal conductivity than those of individual CNTs, GNPs, and simply mixed GNP/CNT fillers, respectively.

2. Experimental

Materials

GNPs and CNTs were purchased from XG Science (grade: M-25, USA) and Nanocyl (grade: NC 7000, USA), respectively. Tin chloride (SnCl₂), palladium chloride (PdCl₂), sodium hypophosphite monohydrate (NaPO₂H₂·H₂O), citric acid, boric acid, sodium hydroxide (NaOH), iron sulfate heptahydrate (FeSO₄·7H₂O), cobalt sulfate heptahydrate (CoSO₄·7H₂O), and hydrochloric acid (HCl, 37%) were purchased from Sigma-Aldrich (USA). Bisphenol A-type epoxy resin (Kukdo Chemical: KFR-120) and amine-based epoxy hardener (Kukdo Chemical: KFH-150) were kindly provided by Kukdo Chemical (Republic of Korea).

Preparation of 3-D CHFs

The procedure to decorate a GNP surface with metal nanoparticles consists of three steps: (1) sensitization, (2) activation, and (3) electroless plating. First, GNPs were sensitized in an acidic SnCl₂ solution (H₂O : HCl = 125 : 1, v/v) using a bath-type sonicator for 60 minutes, and were then washed with deionized water using micro-mesh-sized sieves to remove the unreacted residuals. These sensitized GNPs were activated in an acidic PdCl₂ solution (H₂O : HCl = 400 : 1, v/v) for 60 minutes in a bath-type sonicator, and then similarly washed as mentioned above. The pre-treated GNPs were decorated with monometallic Fe nanoparticles or bimetallic Fe–Co nanoparticles using electroless plating, which was performed in an alkaline solution containing NaOH, FeSO₄·7H₂O (metal precursor 1), CoSO₄·7H₂O (metal precursor 2), NaPO₂H₂·H₂O (reducing agent), citric acid (complexing agent), and boric acid (buffering agent) for 40 minutes at 90 °C. For removing the residuals, Fe- or Fe–Co-decorated GNPs were cleaned with deionized water using micro-mesh-sized sieves.

For directly combining the CNTs with the GNP surface, Fe- or Fe–Co-decorated GNP powders were placed in an open sintered pure alumina boat and thermally treated under an Ar (500 sccm) gas flow in a horizontal quartz tube furnace (70 mm diameter, 1400 mm length) for 20 minutes at 800 °C. A H₂/C₂H₂ (40 sccm/60 sccm) mixed gas flow was applied for 20 minutes at this temperature and then cooled down to room temperature under an Ar (500 sccm) gas flow.

Fabrication of an epoxy/3-D CHF composite

Epoxy composites containing 3-D CHFs were fabricated according to the following steps. First, 20 wt% 3-D CHFs were added to a bisphenol A-type epoxy resin (Kukdo Chemical: KFR-120) and homogeneously mixed with a shear mixer (THINKY Super Mixer: SR-500) for 5 minutes at a speed of 1200 rpm. An amine-based epoxy hardener (Kukdo Chemical: KFH-150) was subsequently added to the epoxy/3-D CHF suspension, and shear mixing was performed for 5 minutes at this same rpm speed. The prepared epoxy/3-D CHF mixture was molded using a hot press (QMESYS heating press tester: QM900M) and cured for 50 minutes at 80 °C. At the gel point, post-curing was additionally performed for 70 minutes at this temperature. The average size of the prepared circular epoxy composites was 12.7 mm (through-plane sample) or 25.4 mm (in-plane sample) in diameter and 1.0 mm in thickness. Since variations of the thickness of the composite may interrupt the reliability of its thermal conductivity, all samples were fabricated without further polishing to ensure that they all had the same thickness. Also, all samples were carefully prepared in order to avoid the formation of pores within and on the surface of the epoxy composite, which greatly affects the thermal conductivity.

Characterization

Surface morphologies of the 3-D CHFs were characterized using a field-emission scanning electron microscope (FE-SEM, Hitachi S-4800, accelerating voltage: 15 kV), and a transmission electron microscope (TEM, JEM-2100F, JEOL). The composition of the surface of the metal-deposited GNPs was analyzed by using energy-dispersive X-ray spectroscopy (EDX, GENESIS 2000, EDAX). The amounts of Fe, Sn, and Pd in the 3-D CHFs were investigated using a thermogravimetric measurement (TGA 2950, TA instrument) system. To identify the crystallinity of the CNTs, Raman spectroscopy measurements (514.5 nm laser, LabRam HR, Horiba Jobin–Yvon) were taken. Thermal conductivity of the epoxy composite was measured using the laser flash method (LFA 447 Nanoflash, Netzsch), which is a standard and popular technique for measuring thermal diffusivities of solid materials above room temperature. Note that additional graphite spray coating was employed on the surface of composite sample to enhance the absorption of Xenon light pulse energy and emission of infrared (IR) radiation by the front and back surfaces of the sample. However, the measured thermal diffusivities depended neither on whether graphite spray coating was performed nor, in the case of our epoxy/3D-CHF composites, on the wt% of the filler-loading (between 20–50 wt%).

3. Results and discussion

Fig. 1(a) shows the experimental procedure for preparing the 3-D CHFs. To directly grow the CNTs on the GNP surface, pristine GNP powders were pre-treated with Pd/Sn components and readily decorated with Fe nanoparticles via electroless plating. After the metal-catalyst-loading on the GNP surface, CNTs spontaneously formed on a GNP surface, and had a downy hair appearance through the CVD process. Fig. 1(b) shows the SEM image of the Fe-decorated GNP surface before CNT synthesis, and the composition of the surface was analyzed using EDX, as shown in Fig. 1(c). From the results of EDX, we identified that GNP surfaces were successfully decorated with Fe nanoparticles via electroless plating. Finally, the 3-D CHFs of a morphology with directly linked junction points between the GNP and CNT were prepared, as shown in Fig. 1(d). To investigate in more detail the junction points between the GNP and CNT, we also took a TEM image (Fig. 1(e)) and found additional direct anchoring points (see inset of Fig. 1(e)). Since the Fe catalyst was situated at the top surface of the CNT, GNP and CNT formed a direct ohmic contact with each other without interlayers; this tip-growth mechanism of the synthesis of CNT is mainly attributed to the weak physical adhesion between the GNP surface and Fe catalyst. During the CVD process, hydrocarbons decomposed on the top surface of the Fe catalyst, and carbons diffused down through the metal. Then, CNT precipitated out across the metal bottom, pushing the whole metal catalyst off the GNP surface.³³ 3-D CHFs prepared from Fe monometallic nanoparticles will be denoted as “3-D CHF:Fe”.

Fig. 1 (a) Experimental procedure for preparing 3-D CHF:Fe. (b) SEM image of an Fe-decorated GNP surface prepared by electroless-plating deposition, and (c) its EDX surface analysis. Surface morphologies of synthesized 3-D CHF:Fe shown by (d) SEM and (e) TEM.

Fig. 2(a)–(c) show the surface morphologies of 3-D CHF:Fe with 15%, 30%, and 55% CNT. Note that these CNT contents on the GNP surface were calculated from the difference in mass of Fe-decorated GNP before and after CNT growth (eqn (1)).

Fig. 2 The 3-D CHF:Fe with (a and d) 15%, (b and e) 30%, and (c and f) 55% CNT. (d–f) Magnified images of 3-D CHF:Fe surface.

In eqn (1), m₀ is the mass of Fe-decorated GNP before CNT growth, and m₁ is the mass of 3D-CHF:Fe after CNT growth. The CNT content on the GNP surface was readily controlled by altering the amount of metal precursor (FeSO₄·7H₂O) during electroless plating. As the amount of metal precursor was increased, more CNTs formed on the GNP surface during the CVD process. Furthermore, the number of CNTs on a single GNP surface proportionally increased with the average CNT diameter (20 nm → 50 nm).

To determine whether defects were present on the CNTs within a 3-D CHF:Fe, a Raman spectroscopic measurement was taken, as shown in Fig. 3(a). For sample preparation, we selectively gathered CNTs via sonication, centrifugation, and extraction of a 3-D CHF:Fe/N-methyl-2-pyrrolidone (NMP) dispersion, and the gathered CNT solution was subsequently spray-coated on a slide glass. As seen in the graph, the CNTs showed D, G, and 2D bands at 1341 cm⁻¹, 1596 cm⁻¹, and 2716 cm⁻¹, respectively, with a calculated I_G/I_D ratio of 0.88. This ratio indicates that CNTs were successfully grown on the GNP surface without severe distortion of their graphitic structures. Furthermore, to determine the relative amount of metal catalysts or impurities (Fe, Sn, Pd) within a 3-D CHF:Fe, we applied TGA to the sample (Fig. 3(b)). The results of this analysis showed the GNPs to include 0.8% impurities, and the 3-D CHF:Fe to include from 3.2% to 6.6% inorganic metal catalysts or impurities, with the higher values resulting from the higher amount of initial metal precursor used.

Fig. 3 (a) Raman spectra of CNTs within a 3-D CHF:Fe. (b) TGA analysis of 3-D CHF:Fe with various CNT contents (15%, 30%, 55%).

With the designed 3-D CHFs:Fe, we investigated thermal conductivity characteristics of epoxy/3-D CHF:Fe composites, as shown in Fig. 4. The epoxy composite was prepared by dispersing the 3-D CHFs:Fe in a bisphenol A-type epoxy resin/amine-based epoxy hardener suspension and by molding the 3-D CHFs:Fe/epoxy/hardener suspension using a hot press. Fig. 4(a) shows the relationship of thermal conductivity (in- and through-plane) versus CNT content within a 3-D CHF:Fe at 20 wt% total filler-loading in an epoxy matrix. In the case of through-plane thermal conductivity, the inclusion of 15–55% CNTs on the GNP surface effectively raised the vertical thermal conductance of the epoxy composite compared with the reference (GNP/epoxy composite, CNT content: 0%), with a maximum vertical thermal conductance for 15% CNT. While the relative amount of CNT on the GNP surface that optimized in-plane thermal conductivity was also 15%, as was observed for the through-plane thermal conductivity, greater amounts of CNTs on the GNP surface hindered horizontal thermal conductance of the epoxy composite in comparison with that of the reference, in contrast to the case of through-plane thermal conductivity. It is believed that numerous CNTs on the GNP surface led to a high prevalence of point-type contact geometry at the neighboring GNP interfaces, resulting in a disturbance of the area of direct contact between the GNPs,²³ and hence the incomplete wetting of epoxy resins between CNTs within a 3-D CHF:Fe, resulting in the formation of interstitial air.³⁴ Thus, we determined that the optimum CNT content within a 3-D CHF:Fe to enhance both in- and through-plane thermal conductivities is approximately 15%. In Fig. 4(b), thermal conductivities of epoxy composites with various carbon fillers, including CNTs, GNPs, simply mixed GNP/CNT, and 3-D CHF:Fe, were directly compared to confirm the degree of enhancement of our 3-D CHF:Fe at 20 wt% total filler-loading. To characterize the thermal conductivity of simply mixed GNP/CNT fillers in the epoxy composite, we used NC 7000 grade CNTs (Nanocyl). As CNTs were incorporated with the GNPs whether they were simply mixed or directly hybridized, through-plane thermal conductivities of epoxy composites were effectively enhanced up to 1.7 times compared to that of the GNP fillers. However, unlike directly hybridized 3-D CHFs:Fe, the in-plane thermal conductivity of the simply mixed GNP/CNT fillers showed a decreased value compared to that of the GNP fillers. This performance reduction is attributed to the density difference between the non-homogenous dispersions of individual GNP and CNT fillers in an epoxy matrix (GNP (2.2 g cm⁻³), CNT (1.7 g cm⁻³)). Consequently, effective thermal conducting pathways were not formed by the CNTs between horizontally-laid GNPs in an epoxy matrix. To investigate the effect of total filler content in an epoxy composite on thermal conductivity, 20–35 wt% 3-D CHFs:Fe (CNT content: 15%) were loaded in an epoxy matrix, and their thermal conductivities were characterized, as shown in Fig. 4(c) and (d). Within the ranges, 3-D CHFs:Fe (CNT content: 15%) effectively enhanced both in- and through-plane thermal conductivities compared to those of GNP fillers. In particular, 3-D CHFs exhibited 1.4 times (@ 20 wt%), 1.1 times (@ 25 wt%), 1.2 times (@ 30 wt%), and 1.1 times (@ 35 wt%) greater in-plane thermal conductivities, and 1.7 times (@ 20 wt%), 1.5 times (@ 25 wt%), 1.3 times (@ 30 wt%), and 1.1 times (@ 35 wt%) greater through-plane thermal conductivities than did the GNP fillers at the same filler loadings. Because of the increase of 3-D CHF:Fe content (20 wt% → 35 wt%) in an epoxy composite, their thermal conductivity enhancements were gradually saturated compared with the reference. And, 3-D CHFs:Fe more effectively increased the thermal conductivity of the epoxy composite at the lower filler content.

Fig. 4 (a) Thermal conductivity of epoxy/3-D CHF:Fe composites with various CNT contents (15%, 30%, 55%) within a 3-D CHF:Fe at 20 wt% total filler-loading. (b) Comparison of thermal conductivities of epoxy composites with various carbon fillers (CNT, GNP, simply mixed GNP/CNT, and 3-D CHF:Fe) at 20 wt% total filler-loading. (c) In- and (d) through-plane thermal conductivities of epoxy/3-D CHF:Fe composites with 15% CNT at various total filler-loadings (20 wt%, 25 wt%, 30 wt%, 35 wt%).

To directly confirm the distribution of 3-D CHFs:Fe within the epoxy composite, the samples were slightly calcined at 400 °C for 5 minutes and observed using SEM. The inset of Fig. 5(a) shows the raw sample of the epoxy composite before calcination, and Fig. 5(a) and (b) show the calcined plane- and cross-sectional views of the sample, respectively. As shown in the figure, CNTs effectively bridged neighboring GNPs in all directions, resulting both in higher in- and through-plane thermal conductivities than those of epoxy/GNP composites at the same filler-loading. This result implies that CNTs effectively bridged neighboring GNPs without disturbing the main orientation and crystallinity of the GNPs and provided additional heat-flowing pathways without severe thermal resistance at the GNP/CNT/GNP interfaces.

Fig. 5 SEM images of calcined epoxy/3-D CHF:Fe composites. (a) Plane and (b) cross-sectional views.

Furthermore, to improve the quality of the CNTs on the GNP surface, bimetallic nanoparticles composed of two ingredients (Fe and Co) were employed on the GNP surface as a catalyst, instead of the monometallic Fe nanoparticles, by taking advantage of electroless-plating deposition. The detailed experimental procedure is shown in Fig. 6(a), and CoSO₄·7H₂O (metal precursor 2) was additionally inserted at the stage of electroless plating in Fig. 1(a). The morphology of 3-D CHFs synthesized from Fe–Co alloy nanoparticles (3-D CHFs:Fe–Co) on the GNP surface is shown in Fig. 6(b) and (c). Unlike the morphology shown in Fig. 2, CNTs on the GNP surface exhibited a straighter morphology than that of the CNTs synthesized from monometallic Fe nanoparticles. This difference in morphology is attributed to the stability of Fe–Co (Fe_0.85Co_0.15) bimetallic catalysts towards oxidation, and carburization reaction avoiding the formation of Fe₃C, which resulted in the bimetallic catalyst being beneficial for the formation of multi-walled CNTs.³⁵ Fig. 6(d) and (e) show the TEM images of synthesized multi-walled CNTs on the GNP surface. With optimized 3-D CHFs:Fe–Co, the thermal conductivities of epoxy composites with 20 wt% total filler loading were characterized, as shown in Fig. 6(f), and directly compared with those of 3-D CHFs:Fe and GNP fillers. Upon the change in the morphology of the CNTs from twisted (Fig. 2(d)) into straight shapes (Fig. 6(c)), the 3-D CHF:Fe–Co displayed a 1.2 times and 1.7 times greater in-plane thermal conductivity and a 1.3 times and 2.1 times greater through-plane thermal conductivity than did the 3-D CHF:Fe (in-plane: 4.7 W mK⁻¹, through-plane: 2.5 W mK⁻¹) and GNP fillers (in-plane: 3.3 W mK⁻¹, through-plane: 1.5 W mK⁻¹), respectively.

Fig. 6 (a) Experimental procedure for fabricating the 3-D CHFs from Fe–Co bimetallic nanoparticles. (b and c) Surface morphologies of 3-D CHF:Fe–Co. (d and e) TEM images of CNTs within a 3-D CHF:Fe–Co. (f) Comparison of thermal conductivities of epoxy composites with GNP, 3-D CHF:Fe, and 3-D CHF:Fe–Co fillers at 20 wt% total filler-loading.

The surface morphology of the calcined epoxy composite containing 3-D CHFs:Fe–Co is shown in Fig. 7(a) (plane view) and 7(b) (cross-sectional view). The straight-shaped CNTs were more effectively linked with neighboring GNPs than those of 3-D CHFs:Fe (Fig. 5). From Raman spectroscopic measurements (Fig. 7(c)), the I_G/I_D ratio was shown to be 1.06, which corresponds to a 20% improvement relative to that of 3-D CHFs:Fe (I_G/I_D = 0.88). As shown in Fig. 7(d), we characterized the thermal conductivity behavior of epoxy/3-D CHF composites with a high total filler content of 46.5 wt% to determine the maximum thermal conductivity of the composite under 50 wt% filler content. Our 3-D CHFs:Fe–Co showed an in-plane thermal conductivity of 30.96 W mK⁻¹ and through-plane thermal conductivity of 4.32 W mK⁻¹ in an epoxy composite; these values correspond to an improvement of 1.5 times for in-plane thermal conductivity and 1.2 times for through-plane thermal conductivity, compared with the those of the GNP fillers.

Fig. 7 Surface morphologies of calcined epoxy/3-D CHF:Fe–Co composites: (a) plane view and (b) cross-sectional view. (c) Raman spectra of CNTs within a 3-D CHF:Fe–Co. (d) Comparison of thermal conductivities of epoxy composites with 3-D CHF:Fe–Co, 3-D CHF:Fe, and GNP fillers at 46.5 wt% total filler-loading.

4. Conclusion

We have demonstrated a 3-D CHF preparation and its application in heat-dissipating epoxy composites. Since electroless-plating deposition enables the simple formation of multi-component metal nanoparticles on an arbitrary substrate, catalytic metal nanoparticles (Fe or Fe–Co), which are beneficial for direct CNT growth, were successfully decorated on the non-reactive pristine GNP surfaces. From catalytic metal nanoparticles, 3-D CHFs were readily synthesized via the CVD process, and their surface morphologies exhibited directly anchored and ohmic-contact structures between GNP and CNT. The 3-D CHF in an epoxy matrix showed efficient heat conduction for both in- and through-plane directions via additional heat-conducting pathways that were formed by the CNTs (nano-sized filler) that bridged between neighboring GNPs (micro-sized filler). This work shows that properly designed three-dimensional carbon-structured fillers via a directly anchored hybridization process may provide a better thermal conducting property within the thermoplastic/thermosetting polymer composites compared with the single carbon filler. Furthermore, our approach offers a convenient, effective, and mass-producible method to hybridize different kinds of carbon nanomaterials and opens up diverse potential applications, ranging from heat-dissipating materials to electrode materials.

Acknowledgements

This study was supported by World Premier Materials (WPM) Program (10037890), and the Industrial Strategic Technology Development Program (10041851) funded by the Ministry of Trade, Industry & Energy (MI) of Korea.

References

1. S. P. Gurrum, S. K. Suman, Y. K. Joshi and A. G. Fedorov, IEEE Trans. Device Mater. Reliab., 2004, 4, 709.
2. J. Wei, Heat Transfer Eng., 2008, 29, 178.
3. V. Singhal, T. Siegmund and S. V. Garimella, IEEE Trans. Compon., Packag., Manuf. Technol., 2004, 27, 244.
4. S. Keeping, LED Heat Dissipation and Lowering Thermal Resistance of LED Lighting Substrates, http://www.digikey.com/en/articles/techzone/2011/aug/led-heat-dissipation-and-lowering-thermal-resistance-of-led-lighting-substrates, accessed: 2, 2011.
5. S. Cho and J. Y. Lee, Electron. Mater. Lett., 2010, 6, 167.
6. R. Prasher, Proc. IEEE, 2006, 94, 1571.
7. R. Siegel and J. Howel, Thermal Radiation Heat Transfer, Tayler & Francis, New York, NY, USA, 4th edn, 2002.
8. Y.-A. Sha, Y.-S. Jhuo, K.-H. Chen and S.-C. Wang, in Microsystems, Packaging, Assembly and Circuits Technology Conference (IMPACT) 2013 8th international (IEEE 2150-5934), IEEE, Piscataway, NJ, USA, 2013, p. 74.
9. Y. Hwang, M. Kim and J. Kim, RSC Adv., 2014, 4, 17015.
10. T.-L. Li and S. L.-C. Hsu, J. Phys. Chem. B, 2010, 114, 6825.
11. E.-S. Lee and S.-M. Lee, J. Am. Ceram. Soc., 2008, 91, 1169.
12. S. Wang, Y. Cheng, R. Wang, J. Sun and L. Gao, ACS Appl. Mater. Interfaces, 2014, 6, 6481.
13. C. Lin and D. D. L. Chung, Carbon, 2009, 47, 295.
14. E. Bekyarova, E. T. Thostenson, A. Yu, H. Kim, J. Gao, J. Tang, H. T. Hahn, T.-W. Chou, M. E. Itkis and R. C. Haddon, Langmuir, 2007, 23, 3970.
15. F. Sarvar, D. C. Whalley and P. P. Conway, in Electronics System Integration Technology Conference (IEEE 1-4244-0553), IEEE, Piscataway, NJ, USA, 2006, vol. 2, p. 1292.
16. R. S. Prasher, J.-Y. Chang, I. Sauciuc, S. Narasimhan, D. Chau, G. Chrysler, A. Myers, S. Prstic and C. Hu, Intel Technol. J., 2005, 9, 285.
17. S. H. Lee, D. H. Lee, W. J. Lee and S. O. Kim, Adv. Funct. Mater., 2011, 21, 1338.
18. A. A. Balandin, S. Ghosh, W. Bao, I. Calizo, D. Teweldebrhan, F. Miao and C. N. Lau, Nano Lett., 2008, 8, 902.
19. S. H. Song, K. H. Park, B. H. Kim, Y. W. Choi, G. H. Jun, D. J. Lee, B.-S. Kong, K.-W. Paik and S. Jeon, Adv. Mater., 2013, 25, 732.
20. K. M. F. Shahil and A. A. Balandin, Nano Lett., 2012, 12, 861.
21. A. A. Balandin, Nat. Mater., 2011, 10, 569.
22. E. Pop, V. Varshney and A. K. Roy, MRS Bull., 2012, 37, 1273.
23. A. Yu, P. Ramesh, X. Sun, E. Bekyarova, M. E. Itkis and R. C. Haddon, Adv. Mater., 2008, 20, 4740.
24. Q.-Q. Kong, Z. Liu, J.-G. Gao, C.-M. Chen, Q. Zhang, G. Zhou, Z.-C. Tao, X.-H. Zhang, M.-Z. Wang, F. Li and R. Cai, Adv. Funct. Mater., 2014, 24, 4222.
25. V. H. Luan, H. N. Tien, T. V. Cuong, B.-S. Kong, J. S. Chung, E. J. Kim and S. H. Hur, J. Mater. Chem., 2012, 22, 8649.
26. W. Feng, J. Li, Y. Feng and M. Qin, RSC Adv., 2014, 4, 10090.
27. A. Dichiara, J.-K. Yuan, S.-H. Yao, A. Sylvestre and J. Bai, J. Nanosci. Nanotechnol., 2012, 12, 6935.
28. W. Li, A. Dichiara and J. Bai, Compos. Sci. Technol., 2013, 74, 221.
29. S. Berber, Y.-K. Kwon and D. Tománek, Phys. Rev. Lett., 2000, 84, 4613.
30. L. Yu, J. S. Park, Y.-S. Lim, C. S. Lee, K. Shin, H. J. Moon, C.-M. Yang, Y. S. Lee and J. H. Han, Nanotechnology, 2013, 24, 155604.
31. L. Yu, H. Kang, Y. S. Lim, C. S. Lee, K. Shin, J. S. Park and J. H. Han, J. Nanosci. Nanotechnol., 2014, 14, 9139.
32. M. Schlesinger and M. Paunovic, Modern Electroplating, Wiley, Hoboken, NJ, USA, 5th edn, 2010.
33. M. Kumar and Y. Ando, J. Nanosci. Nanotechnol., 2010, 10, 3739.
34. A. M. Marconnet, N. Yamamoto, M. A. Panzer, B. L. Wardle and K. E. Goodson, ACS Nano, 2011, 5, 4818.
35. P. Coquay, A. Peigney, E. D. Grave, E. Flahaut, R. E. Vandenberghe and C. Laurent, J. Phys. Chem. B, 2005, 109, 17813.

---