Composites: Part B 58 (2014) 228–234

Contents lists available at ScienceDirect

Composites: Part B

journal homepage: www.elsevier .com/locate /composi tesb

Engineering the coefficient of thermal expansion and thermalconductivity of polymers filled with high aspect ratio silica nanofibers

1359-8368/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.compositesb.2013.10.049

⇑ Corresponding author.E-mail address: ozisik@rpi.edu (R. Ozisik).

Liyun Ren a, Kamyar Pashayi b, Hafez Raeisi Fard b, Shiva Prasad Kotha b, Theodorian Borca-Tasciuc c,Rahmi Ozisik a,⇑a Material Science and Engineering, Rensselaer Polytechnic Institute, Troy, NY 12180, United Statesb Mechanical, Aerospace and Nuclear Engineering, Rensselaer Polytechnic Institute, Troy, NY 12180, United Statesc Biomedical Engineering, Rensselaer Polytechnic Institute, Troy, NY 12180, United States

a r t i c l e i n f o a b s t r a c t

Article history:Received 10 March 2013Received in revised form 23 September 2013Accepted 25 October 2013Available online 6 November 2013

Keywords:A. Nano-structuresA. Polymer–matrix composites (PMCs)B. ThermomechanicalB. Thermal properties

The thermomechanical properties of epoxy filled with two different types of silica nanofillers: sphericalnanoparticles and nanofibers were investigated as a function of silica nanofiller aspect ratio and concen-tration. Results indicated that at room temperature and at 8.74% silica nanofiber concentration (by vol-ume) the thermal conductivity of epoxy increased twofold and coefficient of thermal expansion (CET)decreased by �40%. Silica nanofiber filled epoxy showed 1.4 times greater CET and 1.5 times greater ther-mal conductivity compared to spherical nanoparticle filled epoxy. The significant changes observed inthermomechanical properties of silica nanofiber filled epoxy were attributed to its high aspect ratio byconstraining the polymer matrix as well as reducing the phonon scattering due to the formation of a con-tinuous fiber network within the matrix. In addition to being electrically insulating, the improved prop-erties of silica nanofiber filled epoxy make it an extremely attractive material as underfill andencapsulant in advanced electronic packaging industry.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction use of high aspect ratio fillers [14,15]. To date, no studies were per-

Polymer composites are commonly used as encapsulation andunderfill materials to improve the efficiency of thermal manage-ment in electronic packages. Various inorganic fillers such as silica,alumina, and boron nitride were studied in order to enhance ther-mal conductivity (j) and reduce coefficient of thermal expansion(CTE) of polymeric materials [1–4]. Silica is the most common fillerthat is used to improve thermo-mechanical properties of polymersin microelectronic devices because of its electrical insulation prop-erties, low CTE, good mechanical properties and low cost [5–7].However, in order to achieve the desired thermo-mechanical prop-erties (j > 0.3 W/m K and effective CTE of 20–30 ppm/�C below Tg)[8–10], 55–70% (by volume) of micron-sized silica particles areneeded. The use of such high filler concentrations lead to delami-nation issues during operation and high polymer melt viscosity,which hamper processability of the underfill [11–13]. To optimizethe fabrication and ultimate properties of the electronic packages,it is of great interest to achieve high thermal conductivity (j) andlow CTE (below Tg) with low filler concentration in encapsulationand underfill materials [5].

One possible mechanism of improving j and reduce CTE ofpolymers at low filler loadings (generally less than 10%) is the

formed to study the effect of high aspect ratio silica fillers in elec-tronic packaging applications mainly due to the time consumingnature and the low yield of the high aspect ratio silica fillers syn-thesis [16–18]. In the current study, a facile silica nanofiber syn-thesis method is introduced and thermo-mechanical propertiesof epoxy filled with silica nanofibers and spherical silica nanopar-ticles (from here on referred to as ‘‘silica nanoparticles’’) are stud-ied. It is shown that the use of high aspect ratio silica nanofiberslead to the creation of large confinement regions inside the poly-mer matrix and leads to significant reduction of the coefficient ofthermal expansion. In addition, the percolated network formedby silica nanofibers leads to the enhancement of the thermal con-ductivity. The resulting silica nanofiber reinforced epoxy compos-ites can be used as encapsulation and underfill materials infuture electronic packages.

2. Materials and experimental methods

2.1. Synthesis and characterization of silica nanofibers andnanoparticles

Silica nanofibers used in the current study were fabricated byelectrospinning, utilizing sol–gel precursors. The sol–gel precursorwas prepared by hydrolysis of tetraethyl orthosilicate (TEOS;Sigma–Aldrich) in polyvinyl pyrrolidone (PVP)/ethanol solution.

L. Ren et al. / Composites: Part B 58 (2014) 228–234 229

Usually, one gram of PVP (Mw = 130,000 g/mol; Sigma–Aldrich)was dissolved in 15 g of ethanol (200 proof, Fisher Scientific) andwas stirred at 60 �C for 30 min to form a uniform solution. Then1.5 g of TEOS was added to this polymer solution and after vigorousstirring at 60 �C for 15 min, 0.1 g HCl (2 M) was added to assist thehydrolysis of TEOS. The resulting solution was stirred for another30 min to form the sol–gel precursor. In the electrospinning pro-cess, the sol–gel precursor was loaded into a syringe and was deliv-ered through a metal needle tip by a syringe pump at a constantflow rate of 30 lL/min. The metal needle tip was connected to ahigh voltage supply (20 kV) and a grounded metal collector wasused to collect the final products. The distance between the needletip and the metal collector was fixed at 15 cm. At the end of theelectrospinning process, the synthesized silica/PVP nanofiberswere dried at 80 �C for 6 h. Silica nanofibers were finally obtainedby heat treatment of silica/PVP nanofibers at 650 �C for 12 h.Spherical silica nanoparticles were purchased from Sigma–Aldrichwith an average particle size of 15 nm.

Scanning electron microscope (SEM) Supra 55 (Zeiss, Germany)was used to characterize the morphology and diameter of silicananofibers. Silica nanoparticles were characterized with transmis-sion electron microscope (TEM) JEOL 2010 (JEOL, Japan) under200 kV. FTIR spectra of silica nanoparticles and nanofibers wereobtained from a Perkin Elmer Spectrum One FTIR spectrometer(Waltham, MA). Potassium bromide (KBr) powder was first driedat 300 �C for 2 h, and then was blended with silica nanofillers be-fore testing. The resulting mixture was pressed into pellets foranalysis. Thirty-two scans were averaged to obtain the final signal.The signal was background corrected. The X-ray diffraction (XRD)patterns were recorded on a Panalytical X’Pert diffractometer(Westborough, MA) using Cu Ka radiation with a nickel filter. Silicananoparticles were placed in a zero background silicon holder forXRD pattern scanning. Silica nanofibers were attached to a dou-ble-sided tape and the final XRD pattern was obtained by subtract-ing the background pattern of the double-sided tape.

2.2. Synthesis and characterization of silica/epoxy nanocomposites

Silica/epoxy composites were prepared through a solvent evap-oration process. Huntsman Araldite epoxy system, which includesbisphenol A diglycidyl ether liquid epoxy resin, aliphatic/aromaticacid anhydride hardener, and liquid tertiary amine catalyst, wasused in the current study. Silica nanofillers and epoxy resin werefirst mixed with acetone. After the evaporation of the solvent at80 �C under vacuum, epoxy/silica mixtures were mixed with hard-ener (weight mixing ratio of resin to hardener was 1:1) and accel-erator (1% by weight). High shear mixing was used to assist furtherdispersion of silica nanofillers in epoxy. Silica/epoxy mixtures werethen degassed at 60 �C under vacuum. After degassing, sampleswere first cured at 80 �C for 10 h and then post-cured at 140 �Cfor 10 h.

SEM was used to characterize the distribution of silica nanofil-lers in epoxy. Before characterization with SEM, platinum coatedsamples were processed with a focused ion beam (FIB) nanolithog-raphy to mill a square hole or a cross-section. The FIB milled sur-faces are shown in Fig. 1. Fig. 1a shows a typical FIB milledsurface used for the characterization of silica nanoparticle filledepoxy composite. Cross-sectioned surfaces were generated for sil-ica nanofiber filled epoxy because the length of nanofibers is on theorder of micrometers; therefore, various cross-sections with differ-ent sizes were made according to the concentration of silicananofibers.

Silica nanofiber distribution in epoxy was further analyzed bytransmission optical microscope (TOM) images. For optical imageanalysis, three or four specimens were prepared at each silica con-centration by polishing sample surfaces. For each specimen, ten

images were taken at different optical focal planes. All the opticalimages were studied with Image J to measure silica nanofiberlength and orientation distribution.

2.3. Coefficient of thermal expansion measurements

Coefficient of thermal expansion was measured with SeteramTMA-90 thermo-mechanical analyzer according to ASTM E831-06[19]. Three samples were tested at each silica concentration. Dur-ing testing, the displacement change of a 5 mm long cylindricalsample was recorded from 40 to 90 �C. Final effective CTE belownanocomposite glass transition was calculated with the aid of Set-soft analysis software of TMA-90.

The measured CTE of the epoxy/silica nanocomposites wascompared to a mean field model suitable for composites with3-dimensional fillers with random distribution within a matrix[20–22]. The model used in the current work was based on amean field theory with misaligned inclusions [20] by utilizingthe work of Takao and Taya [21]. The theoretical model requiresfiller aspect ratio for the estimation of the CTE of the compositesystem. A value of 260 was used as the aspect ratio of silicananofibers in all theoretical calculations. This value of the aspectratio was obtained from measurement of silica nanofibers and isprovided in Section 3.

2.4. Thermal conductivity measurements

The experimental setup used to measure the thermal conduc-tivity of composites is shown in Fig. 2. Thermal conductivity test-ing setup consists of a heater, a thin foil of indium, sample,another thin foil of indium, and a heat sink at the bottom. Twothermocouples were embedded in the indium foils to read the tem-perature across the sample. First, total thermal resistivity was cal-culated using Rs+int = DT/Q, where Rs+int is the total thermalresistivity of the sample (Rs) and interface thermal resistance (Rint),Q is the total heat flux generated by the heater, and DT is the tem-perature difference across the sample. This value of the total ther-mal resistivity was used to plot A � Rs+int vs. sample thickness,where A is the sample area. Rint is obtained from the intercept ofA � Rs+int vs. sample thickness plot (at zero thickness). In addition,the reciprocal of the slope of the A � Rs+int vs. sample thickness plotprovides the j of the sample. All tests were performed according toa modified version of ASTM D5470 [19]. Each data point representsthree to four measurements. The scatter observed in the j data wasless than 5%.

Ordonez-Miranda model and the two-phase Lewis–Nielsenmodel were used to fit the experimental data. Ordonez-Mirandamodel is described as follows [23]:

j=jm � a1� a

� �jm

j

� �1=1þr¼ 1� f ð1Þ

In this equation j is the effective thermal conductivity of thecomposite, jm is the thermal conductivity of the matrix, f is thevolume concentration, and r is the filler shape factor. For cylindri-cal particles r = 1. In Eq. (1), a is defined as follows:

a ¼ jm

jp� aK

a

� ��1

ð2Þ

where jp is the particle thermal conductivity, a is the particle ra-dius, and aK is the Kapitza radius (aK = R � jm), and R is the interfacialthermal resistance. In the current study, a value of 30 nm was ob-tained for aK from the best fit to thermal conductivity vs. silica con-centration, and this value is within the accepted range of Kapitzaradius values reported elsewhere [23,24]. Eq. (2) can be simplifiedfor spherical particles as follows:

Fig. 1. Typical FIB surfaces for (a) silica nanoparticle filled epoxy and (b) silica nanofiber filled epoxy.

Fig. 2. Schematics of the thermal conductivity measurement setup.

Table 1The fitting parameters used in the Lewis–Nielsen model calculations.

Composite jm (W/m K) n /f (W/m K) /M

Epoxy–SiO2 nanoparticle 0.193 1.50 1.36 0.637Epoxy–SiO2 nanofiber 0.193 260 1.36 0.520

230 L. Ren et al. / Composites: Part B 58 (2014) 228–234

jjm¼ aþ Uþ þ U� ð3Þ

where U is expressed as follows:

U� ¼ ba2

1� 1� 4b3

27a2

!1=224

35

8<:

9=;

1=3

ð4Þ

where

b ¼ ð1� aÞð1� f Þ ð5Þ

The two-phase Lewis–Nielsen model for predicting compositethermal conductivity is expressed as follows [25]:

j ¼ jm

1þ n/fjf � jm

jf � njm

� �

1� /fjf � jm

jf � njm

� �1þ /f

1� /M

/2M

!" #8>>>><>>>>:

9>>>>=>>>>;

ð6Þ

In Eq. (6), n is a constant that depends on the shape, orientationand aspect ratio of the dispersed filler. The quantity /M is the max-imum packing factor of the filler. This model assumes zero interfa-cial thermal resistance. All the fitting parameters used in thismodel are tabulated in Table 1.

3. Results and discussion

3.1. Characterization of silica nanofibers

Fig. 3 presents the scanning electron microscope (SEM) image ofsilica–PVP nanofibers (Fig. 3a), SEM image of silica nanofibers afterheat treatment (Fig. 3b), transmission electron microscope (TEM)image of silica nanoparticles (Fig. 3c), and the diameter distribu-tion of silica nanofibers (Fig. 3d). As described previously, thesilica nanofibers were obtained from the heat treatment ofPVP–silicaChange "PVP-silica" to "silica-PVP". nanofibers. Afterthe removal of PVP template, the average silica nanofiber diameterdecreased to 70 nm. Silica nanoparticles have a very narrow sizedistribution and the average diameter of silica nanoparticles wasmeasured as 15 nm.

Silica nanofibers and silica nanoparticles were characterized bythe Fourier transformation infrared spectroscopy (FTIR) as shownin Fig. 4. It was found that both types of fillers only show typicalSi–O bond vibrations as observed by the peaks at 800, 1123, and1220 cm�1 [26].

Fig. 5 shows the X-ray diffraction (XRD) analysis of both silicananoparticles and silica nanofibers to verify the crystal structureof both fillers. The broad peak around 20� in both cases confirmsthe amorphous structure of the silica nanoparticles and silicananofibers used in the current study [27].

3.2. Characterization of epoxy/silica nanocomposites

Fig. 6 shows the nanoparticle distribution in epoxy. As illus-trated in Fig. 6, the nanoparticles are uniformly dispersed inthe epoxy matrix with limited amount of agglomeration. Theaverage size of silica nanoparticle agglomerate was found to be�44 nm as shown in Fig. 6a. Fig. 6b was generated by analyzingthe SEM images of the nanocomposites with the Image Jprogram.

Fig. 7 shows representative images of silica nanofiber distri-bution in epoxy. It is shown that at low filler concentrations, sil-ica nanofibers have an average length of 45 lm, which yields anaverage aspect ratio of approximately 260. This average aspectratio was used in Table 1 as the n value. The orientation distri-bution analysis showed that silica nanofibers were distributedrandomly at the studied filler loading ranges up to 8.74%. Thesilica nanofiber diameters reported in Fig. 7d are different thanthose shown in Fig. 3d because in Fig. 7d, the nanofibers are

Fig. 3. (a) SEM image of electrospun silica–PVP nanofibers, (b) SEM image of silica nanofibers after heat treatment of silica–PVP nanofibers, (c) TEM image of silicananoparticles, and (d) diameter size distribution of silica nanofibers as obtained from SEM image analysis.

0

20

40

60

80

100

120

400 1200 2000 2800 3600

Tra

nsm

ittan

ce (

au)

Wavenumber (cm-1)

Silica nanoarticle

Silica nanofiber

Fig. 4. FTIR spectra of silica nanofibers and silica nanoparticles.

10 20 30 40 50

Inte

nsity

(au

)

2θ (degrees)

Silica nanoparticle

Silica nanofiber

Fig. 5. XRD patterns of silica nanofibers and silica nanoparticles.

L. Ren et al. / Composites: Part B 58 (2014) 228–234 231

coated with epoxy matrix, thereby, a greater average size is ob-tained. The fact that the silica nanofibers are covered by epoxysuggests that there is a strong interaction between silica andepoxy.

3.3. Coefficient of thermal expansion of silica/epoxy nanocomposites

Fig. 8a presents coefficient of thermal expansion (CTE) of silicananofiber filled epoxy as a function of silica concentration. For bothtypes of silica fillers, the thermal expansion coefficient decreasedwith increasing silica concentration. At 8.74% silica nanofiber con-tent, the thermal expansion coefficient was observed to be34 ppm/�C, whereas to achieve the same CTE value epoxy mustbe filled with 40% micron-sized silica particles [11,28,29]. Silicananofiber filled epoxy exhibited lower CTE than silica nanoparticlefilled epoxy at all filler concentrations studied. High aspect ratio(nanofiber) filled epoxy showed a significant reduction in CTE. This

is because the high aspect ratio fillers can provide remarkablemechanical constraint to the deformation of the polymer matrixby interfering with the thermal stress distribution inside the ma-trix material. In order to quantify the CTE reduction accountingfor high aspect ratio fillers, a mean field model for randomlyoriented fiber inclusions was used to fit the experimental data.The CTE of silica nanofiber filled epoxy was approximately 30%lower than that predicted by the model. Of course, the mean fieldmodel is not designed for nanometer sized fillers, therefore the dis-crepancy between the model and the experimental findings are notsurprising. It was previously shown that the mean field model doesa good job for systems with micron-sized fillers [30]. However,there are other effects that lead to the discrepancy observed be-tween experimental findings and the model prediction. Firstly,the CTE of composites is influenced by the number of fillers, whichconstrain the deformation of the matrix. Because the amount of ri-

Fig. 6. (a) SEM image of 8.74% silica nanoparticle filled epoxy composite from FIB surface, (b) silica nanoparticle size distribution.

Fig. 7. (a) TOM image of 2.8% silica nanofiber filled epoxy; (b) silica nanofiber length distribution; (c) SEM image of 8.74% silica nanofiber in epoxy; (d) silica nanofiberdiameter distribution in epoxy.

Fig. 8. (a) CTE change of silica nanofiber filled epoxy below Tg and (b) representation of the constraining effect of nanofillers on the epoxy matrix as depicted by gray shadedregions as a function of filler size and aspect ratio (objects are not drawn to scale).

232 L. Ren et al. / Composites: Part B 58 (2014) 228–234

L. Ren et al. / Composites: Part B 58 (2014) 228–234 233

gid fillers per unit volume increases tremendously with decreasingfiller size (the number of fillers increased �109 times from micron-sized fillers to nano-sized fillers), nano-sized fillers exhibitsubstantially greater interfacial region between the filler and thematrix. These interfacial regions provide significant physical andmechanical constraint to the matrix (as illustrated in Fig. 8b). Sec-ondly, the strong deformation constraint effect of high aspect ratiofillers becomes quite significant in the case of nano-sized fibers. Itis possible that the existence of high aspect ratio fillers (comparedto spherical fillers) act as highly effective obstacles to chain relax-ation similar to entanglements. These physical obstructions areless relevant in the case of spherical fillers as the polymer chainshave easier time relaxing around the smaller spherical filler. As aresult, silica nanofiber filled epoxy exhibited the lowest CTE com-pared to nanoparticle filled epoxy.

3.4. Thermal conductivity of silica/epoxy nanocomposites

Another criterion for efficient electronic package thermal man-agement is to increase the thermal conductivity of the epoxy. Fig. 9shows the thermal conductivities of silica nanoparticle and nanofi-ber filled epoxy at various filler concentrations. Best fits by two-phase Lewis–Nielsen and Ordonez-Miranda models are also pre-sented in Fig. 9. The fitting parameters are provided in Table 1.As illustrated in Fig. 9, the thermal conductivity of epoxy increasedfor both silica nanoparticle and nanofiber filled epoxy as a functionof silica concentration. The thermal conductivity of silica nanopar-ticle filled epoxy is generally lower than that of silica nanofiberfilled epoxy. The maximum thermal conductivity is found to be0.37 W/m K at 8.74% silica nanofiber concentration. In order toachieve the same thermal conductivity in epoxy with micron-sizedsilica, silica concentration of 50% was required [31,32,8].

Usually, a percolation threshold as a function of concentration isobserved in thermal conductivity behavior. However, a percolationthreshold was not observed in Fig. 9. This is because of the closeproximity of the thermal conductivities of the silica (1.38 W/m K)and epoxy (0.19 W/m K) to each other (jf/jm � 7). Normally athermal conductivity ratio on the order of 1015 is needed to ob-serve drastic changes in the thermal conductivity behavior of com-posites [33].

To understand the significant difference of thermal conductivityenhancement in various systems, the influence of filler/matrixinterface and aspect ratio need to be considered. Both nanofibers

0.15

0.20

0.25

0.30

0.35

0.40

0 2 4 6 8 10

The

rmal

Con

duct

ivity

(W

/m.K

)

Silica Concentration (%)

Neat Epoxy Silica nanoparticle filled epoxy Silica nanofiber filled epoxy Ordonez-Alvarado model Lewis-Nielsen model

Fig. 9. Thermal conductivities of nanoparticle and nanofiber silica filled epoxy.Ordonez Alvarado model was calculated for a spherical particle and Lewis–Nielsenmodel was calculated for a fiber with an aspect ratio of 260.

and nanoparticles provide large interfacial regions, which resultsin large interfacial thermal resistance [34,35]. Interfacial thermalresistance produces high levels of phonon scattering and impedeseffective heat transfer in the composite. To find the interfacial ther-mal resistance of silica/epoxy interface, the Ordonez-Mirandaeffective thermal conductivity model [36] was implemented. Thefit of Ordonez-Miranda model to experimental data resulted inan interfacial thermal resistance of 1.5 � 10�8 m2 K/W between sil-ica nanoparticles and epoxy. Because the interface density in nano-particle filled epoxy is greater than that in nanofiber filled epoxy,nanoparticle filled composites should yield a greater interfacialthermal resistance compared to nanofiber filled composites [25].Furthermore, silica nanofibers with their high aspect ratio (�260)and micron-sized lengths (30–50 lm) could form a physically con-nected, thermally conductive percolated network. This continuousfiber network effectively reduces phonon scattering in silica nano-fiber filled epoxy and provides a more efficient thermal conductingcapability than the scattered silica nanoparticle clusters. The ab-sence of significant interfacial thermal resistance increase in nano-particle filled epoxy can, therefore, be attributed to a lack of apercolated network. In order to understand the influence of filleraspect ratio effect, Lewis–Nielsen model was used to fit the silicananofiber filled epoxy experimental data. The filler aspect ratio of260 that was obtained from TOM images was used in the Lewis–Nielsen calculations. It is important to note that the Lewis–Nielsenmodel takes into account the effects of filler size and geometry butneglects interface thermal resistance [23,25]. Comparison of theLewis–Nielsen model and experimental results from silica nanofi-ber epoxy showed a 95% agreement, which indicates that interfa-cial thermal resistance has little influence on the thermalconductivity of silica nanofiber/epoxy nanocomposite.

4. Conclusions

In conclusion, enhancements in the thermal conductivity andcoefficient of thermal expansion were observed in silica nanofiberfilled epoxy compared to silica nanoparticle filled epoxy, neatepoxy, and literature reported micron-sized silica filled epoxy.The increased thermal conductivity and reduced coefficient ofthermal expansion were achieved at low filler concentrations com-pared to both spherical silica nanoparticle filled and micron-sizedsilica filled epoxy. It was shown that the high aspect ratio silicananofibers are highly effective in improving epoxy thermo-mechanical properties, and the high aspect ratio filled epoxy hasthe potential to be used as a new generation of high-efficiencyelectronic encapsulation and underfill material.

Acknowledgements

This material is partially based upon work supported by IBMand the National Science Foundation under Grant Nos. 1200270and 1003574.

References

[1] Zhou W, Qi S, An Q, Zhao H, Liu N. Mater Res Bull 2007;42:1863–73.[2] Lee G, Park M, Kim J, Lee J, Yoon H. Compos A Appl Sci Manuf 2006;37:727–34.[3] Pilling MW, Yates B, Black MA, Tattersall P. J Mater Sci 1979;14:1326–38.[4] Kidalov SV, Shakhov FM, Vul AY. Diam Relat Mater 2008;17:844–7.[5] Lu D. Materials for advanced packaging. US: Springer; 2009.[6] Haleh Ardebili MP. Encapsulation Technol Electron Appl 2009.[7] Merrill AIHC, Minges L. Electronic materials handbook: packaging, 1989.[8] Wong CP, Bollampally RS. J Appl Polym Sci 1999;74:3396–403.[9] Sun Lee W, Yu J. Diam Relat Mater 2005;14:1647–53.

[10] Chen Y, Lin H. J Polym Res 2003;10:247–58.[11] Wong CP, Bollampally RS. Mater Sci 1999;22:54–9.[12] Kang S, Hong SI, Choe CR, Park M, Rim S, Kim J. Polymer 2001;42:879–87.[13] Licari JJ. Adhesives technology for electronic applications: materials,

processing, reliability. 2nd ed. US: William Andrew; 2011.

234 L. Ren et al. / Composites: Part B 58 (2014) 228–234

[14] Li J, Ma PC, Chow WS, To CK, Tang BZ, Kim JK. Adv Funct Mater2007;17:3207–15.

[15] Wang S, Tambraparni M, Qiu J, Tipton J, Dean D. Macromolecules2009;42:5251–5.

[16] Mellado P, McIlwee HA, Badrossamay MR, Goss JA, Mahadevan L, Parker KK.Appl Phys Lett 2011;99:203107.

[17] Li D, Zhao L, Jiang C, Lu JG. Nano Lett 2010;10:2766–71.[18] Lunkenbein T, Breu J. Chem Mater 2012;24:1802–10.[19] ASTM Annual Book of Standards, ASTM, International, 2008.[20] Doghri I, Tinel L. Comput Methods Appl Mech Eng 2006;195:1387–406.[21] Takao Y, Taya M. J Appl Mech 1985;52:806–10.[22] Mallick K, Cronin J, Arzberger S. Composite Technology Development Inc.,

Lafayette Co., 2006; 47th AIAA/ASME/ASCE/AHS/ASC Structures, StructuralDynamics, and Materials Conference, Newport, Rhode Island, May 1–4, 2006.

[23] Ordonez-Miranda J, Alvarado-Gil JJ. Compos Sci Technol 2012;72:853–7.[24] Nan C-W, Liu G, Lin Y, Li M. Appl Phys Lett 2004;85:3549–51.[25] Nielsen LE. Ind Eng Chem Fund 1974;13:17–20.

[26] Beganskiene A, Sirutkaitis V, Kurtinaitiene M, Juškenas R, Kareiva A. Mater Sci2004;10:287–90.

[27] NIOSH Manual of Analytical Methods, 4th Edition Third Supplement: DHHS(NIOSH) Publication No. 2003-154, Center for Disease Control and Prevention,<http://www.cdc.gov/niosh/docs/2003-154/pdfs/7501.pdf>.

[28] Chaturvedi M. Acta Metall 1998;46:4287–302.[29] Suzuki N, Kiba S, Yamauchi Y. Phys Chem Chem Phys 2011;13:4957–62.[30] Paul DR, Fornes TD, Yoon PJ. Polymer 2002;43:6727–41.[31] Lee W, Han I, Yu J, Kim S, Byun K. Thermochim Acta 2007;455:148–55.[32] Li H, Jacob KI, Wong CP. IEEE Trans Adv Packag 2003;26:25–32.[33] Song YS, Youn JR. Carbon 2005;43:1378–85.[34] Shenogin S, Xue LR, Ozisik R, Keblinski P, Cahill DG. J Appl Phys

2004;95:8136–44.[35] Huxtable ST, Cahill DG, Shenogin S, Xue L, Ozisik R, Barone P, et al. Nat Mater

2003;2:731–4.[36] Ordonez-Miranda J, Yang R, Alvarado-Gil JJ. Appl Phys Lett 2011;98:

233111.