Composites: Part B 44 (2013) 698–703

Contents lists available at SciVerse ScienceDirect

Composites: Part B

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

Thermal properties of aluminum–graphite composites by powder metallurgy

J.K. Chen ⇑, I.S. HuangInstitute of Materials Science and Engineering, National Taipei University of Technology, Taipei 10608, Taiwan, ROC

a r t i c l e i n f o

Article history:Received 12 November 2011Received in revised form 15 January 2012Accepted 24 January 2012Available online 2 February 2012

Keywords:A. Metal–matrix composites (MMCs)B. AnisotropyB. Thermal propertiesE. Sintering

1359-8368/$ - see front matter � 2012 Elsevier Ltd. Adoi:10.1016/j.compositesb.2012.01.083

⇑ Corresponding author. Address: 1, Sec. 3, ZhonTaiwan, ROC. Tel.: +886 2 2771217x2763; fax: +886 2

E-mail address: jkchen@ntut.edu.tw (J.K. Chen).

a b s t r a c t

10–90 vol.% of flake graphite is combined with aluminum powders to form aluminum–graphite compos-ites via vacuum hot pressing process. The results show that the relative density of powder compositesachieves up to 99% for composites containing 10–90 vol.% graphite. With the increase of flake graphitefrom 10 to 90 vol.% in the composites, thermal conductivity increase from 324 to 783 W/m K; the coef-ficients of thermal expansion (CTE) in direction parallel to basal planes of graphite flakes decrease from16.9 to �2.5 ppm/K, and the CTE perpendicular to basal plane decreases from 15.2 to 10.1 ppm/K. Theparallel arrangements of graphite flakes and aluminum lead the thermal conductivity values of compos-ites to agree well with parallel model estimations. The CTE of composites in a-axis agree with rule of mix-ture estimation representing proportional contribution of the two phases. Along the c-axis of graphite incomposites, the CTE are consistent with Turner model estimations indicating that the graphite flake couldabsorb the expansion from the aluminum phase.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Thermal management material has been of significance in theelectronic industry [1]. The development of electronic devices withhigher calculating speeds within a more compact size leads tomore heat generation per device. In order to meet thermal dissipa-tion requirements for electronic packaging applications, materialswith high thermal conductivity and low thermal expansion mustbe developed [2].

Traditional electronic packaging materials are mostly pure met-als that have excellent thermal conductivity, e.g. Cu, Al, Au and Ag.However, these metals usually also have high coefficients of ther-mal expansion which are much higher than that of semiconductordevice to be packaged. Besides, high density and costs are alsoproblems for the applications of these metals. Recently, develop-ment of ceramic (e.g. BeO [3], AlN [4], Al2O3 and SiC [5]) and metalmatrix composites (e.g. Cu–Mo and Cu–W [6]) has been quicklyreplacing the uses of monolithic metals for thermal dissipation inelectronic packaging [7]. Among these, Al-based metal matrix com-posites, such as Al–SiC [8] and Al–graphite [9], has matched coeffi-cient of thermal expansion and good mechanical properties, butthe poor metallization of filler phase could lead to heat damage[10] due to thermal mismatch strain [11].

The pressure infiltration of the liquid metal into the ceramicpreform was developed [12,13] to improve the wettabilitybetween filler and matrix alloy [14]. Composites containing higher

ll rights reserved.

g-Xiao E. Rd., Taipei 10608,27317185.

volume percentage of fillers can also be fabricated using liquid-based process. However, Al4C3 phase is often found to form onthe fillers by reacting with liquid metals in either aluminum–graphite [15] or aluminum–SiC [8] composites. The Al4C3 phaseis detrimental to the composites by greatly reducing both theirthermal conductivity and mechanical properties.

To avoid the formation of Al4C3 phase, powder metallurgy routeis employed in current study to fabricate Al–graphite compositesinstead of conventional liquid-based process. Although the charac-teristic plate-shaped graphite could make it a less ideal filler frommass production point of view, proper molding can be designedto align the graphite flakes. The anisotropic physical propertiescan also be taken advantage by aligning the basal planes of graphiteflakes in direction consistent with heat transfer. For example,pulsed electric current sintering process was employed to makealuminum–graphite fiber composites with thermal conductivityas high as 600 W/m K [16] and coefficient of thermal expansionbelow 20 ppm/K. Therefore, high thermal conductivity could beachieved in the same time while the coefficient of thermal expan-sion is controlled.

2. Experimental procedures

Commercially available aluminum powders (99.9% purity;d50 = 35 lm) and natural flake graphite (graphitization degree98%; d50 = 550 lm and thickness = 10–30 lm) were mixed usingdry process. 10–90 vol.% of flake graphite is added in the compositematerials. The flake graphite has its surface parallel to the basalplane and thickness parallel to the c-direction as shown in Fig. 1.

Fig. 1. XRD analysis of flake graphite showing flake surface parallel to the basalplane.

J.K. Chen, I.S. Huang / Composites: Part B 44 (2013) 698–703 699

The mixtures of aluminum and flake graphite were filled in agraphite mold for hot press. The flake graphite stacks naturallyby layer into the mold with tapping while filling. Hot press processwas then performed at 913 K and a pressure of 60 MPa for 1 hunder 2 � 10�2 torr vacuum. After hot pressing, different samplescan be cut from desired directions so that thermal properties aremeasured along interested directions.

Microstructures of sintered materials were examined with opti-cal microscopy, scanning electron microscopy (SEM), and X-ray dif-fraction (XRD). And the density of composite was measured usingArchimedes method. Aluminum–graphite composite disks (£2.7 mm � 3 mm) were carbon-coated for thermal diffusivity mea-surement. Thermal diffusivity was measured via laser flash tech-nique using a NETZSCH LFA447 NanoFlashTM thermal constantanalyzer. The measurements were performed such that orientationof heat dissipation was parallel to the basal plane of graphite flakes.The heat capacity of composite was measured by differential scan-ning calorimeter (Perkin Elmer Pyris) at 300 K. Thermal conductiv-ity is thus obtained by multiplying composite density, specific heat,and thermal diffusivity of the composites. Coefficients of thermalexpansion were measured from room temperature to 373 K usingNETZSCH DIL 402C dilatometer and the shape of the specimenwas 10 mm � 5 mm � 5 mm.

3. Results and discussion

3.1. Microstructures

Fig. 2 shows the microstructures of Al–graphite compositeswith different graphite percentage. The brighter phase is alumi-num and dark region is nature flake graphite. No significant poresexist in the interface between aluminum and graphite in Fig. 2a–gor composites containing up to 70 vol.% graphite. The use of hotpressing temperature close to melting point of aluminum andpressure under vacuum atmosphere apparently provide adequatewettability for graphite to sinter with aluminum.

Increased pores can be observed in samples containing higherthan 80 vol.% graphite or Fig. 2h–i. With increasing graphite, thebonding phase or aluminum decreases and becomes harder toinfuse among graphite flakes. Furthermore, it is observed that thegraphite flakes are bent severely in Fig. 2i containing 90 vol.%graphite. The flakes are wavy due to the stacking of large amountof graphite flakes with different size under the pressure. The stack-ing of graphite also closes the passage for aluminum to diffusebetween the flakes. Larger pores are thus formed at the graphiteinterfaces. These pores apparently will affect the density, thermalproperties, and mechanical properties of composites.

Etter et al. [15] reported that aluminum and graphite interfacemay form needle-like Al4C3 phase at aluminum–graphite interfacevia liquid infiltration process. The Al4C3 phase is brittle and tendsto absorb moisture which degrades both the mechanical propertiesand thermal conductivity of composites. In this study, Al4C3 wasnot observed (Fig. 3). The absence of the Al4C3 phase is alsoconfirmed by XRD analysis in Fig. 4, while the graphite peakincreases with amount of flake graphite.

There are two reasons that the current process suppressesformation of Al4C3 phase. Mizuuchi et al. [17] demonstrated thatsurface damages of fillers by direct contact with molten aluminumcause the reductions of materials properties. The hot press temper-ature of 913 K is lower than the melting point of aluminum(933 K). The formation of Al4C3 can thus be avoided. Furthermore,flake graphite with nearly 100% graphitization degree [18] isemployed in current study which is higher than those used inthe literature [15,19]. The more stable flake graphite with largerflat surface has lower tendency for Al4C3 phase formation.

Fig. 5 shows the theoretical and experimental density of alumi-num–graphite composites. The density apparently decreases line-arly with increase of graphite content except at graphite contentmore than 80 vol.%. Fig. 6 further shows that the relative densityof composites lies mostly at 97–99%. From both Figs. 2 and 3,aluminum and graphite appears to have very good wetting at theinterface. Only very limited amount of pores are observed mostlyat the graphite:graphite interface. The constant <3% open porosityas shown in Fig. 6 could be caused by the pores within the graphiteflakes or graphite damage in sample preparation process.

When the graphite content is higher than 80 vol.%, the theoret-ical density decreased to below 95%. The pores are observed moreeasily in Fig. 2h and i. In these composites, aluminum diffuses withdifficulty among the greater number of stacked graphite flakes andresults in much higher percentage of pores. From percolationpoints of view, this indicates that the graphite flakes form closenetworks which hinders the infusion of aluminum. The closednetworking of graphite flakes are especially significant whengraphite content is higher than 80 vol.%.

3.2. Thermal conductivity of Al–graphite composites

Thermal conductivity of composites is obtained by multiplyingdensity, specific heat and heat diffusivity as described in Section2. The thermal conductivity of Al–graphite composites with differ-ent graphite content are shown in Fig. 7. It is noted that the graphiteflakes are arranged such that their basal planes are parallel to thedirection of thermal conduction measurement. Therefore, withgraphite content increasing from 10 to 80 vol.%, thermal conductiv-ity of composites increases from 324 W/m/K to 783 W/m/K. Thesevalues are all much higher than that of pure aluminum (237 W/m/K) due to the high thermal conductivity of graphite along thethermal conducting direction, or in current study, the basal planesof graphite flakes. These values are also higher than those reportedby Chang et al. [20] with similar percentage of graphite. Size ofgraphite flakes and difference in process could combine to causethe difference. In previous study [20], 100 lm graphite and squeezecasting process is used. As described in Section 3.1, liquid-basedprocess could degrade the aluminum–graphite interface. The larger550 lm graphite flakes used in current study are also advantageousfor both a smaller interface thermal resistance and improved con-nections among the flakes. The networks by connecting graphiteflakes provide continuous passages for heat transfer. The high ther-mal conductivity achieved in Al–80 vol.% graphite composite con-firms the good networking among graphite, even though itsrelative density is below 95%. The conductive ‘‘channel’’ of fillersare also responsible for thermal conduction in a recent work by

Fig. 2. Optical micrographs of Al–graphite composites with (a) 10, (b) 20, (c) 30, (d) 40, (e) 50, (f) 60, (g) 70, (h) 80, and (i) 90 vol.% graphite flakes taken along the basal planes.

Fig. 3. SEM micrographs showing no Al4C3 formation in Al–graphite compositescontaining 50 vol.% graphite (aluminum: lighter region; graphite: dark region). Fig. 4. XRD spectra of aluminum–graphite composites with different graphite

contents taken with sample surfaces parallel to the flake surface planes.

700 J.K. Chen, I.S. Huang / Composites: Part B 44 (2013) 698–703

Baglari et al. [21] on thermal conductivity of flash ash-polymercomposites.

It is further noted that these values are no lower than diamond/copper composites reported by Yoshida and Morigami [22] whereup to 80 vol.% of diamond is added into copper demonstrating

742 W/m/K thermal conductivity. In the same study, diamond oflarger particle size is also shown to increase thermal conductivity.

On the other hand, in composite containing 90 vol.% graphite,thermal conductivity drops to 774 W/m/K due to larger amount

Fig. 5. Theoretical and measured density of aluminum–graphite composites.

Fig. 6. Relative density and open porosity of aluminum–graphite composites.

J.K. Chen, I.S. Huang / Composites: Part B 44 (2013) 698–703 701

of pores along graphite flake surface. Apparently the greaterincrease of porosity outweighs the increase of graphite. The poresare formed due to the lack of aluminum binding among neighbor-ing graphite which hinders the transfer of heat. Furthermore, thegraphite flakes are lack of mechanical strength and are easily bentto deform which could also deflect flow of heat or even damage theflakes as shown in Fig. 2i.

Theoretical calculations are made using parallel model. Parallelmodel consider thermal conductivity of composites with parallellayers of alternating lamellae [23]. Such alternating lamellae arerepresented nearly ideally in the case of Al–graphite composites.The effective thermal conductivity, Keff, can be expressed as:

Keff ¼ V1K1 þ V2K2 ð1Þ

where K1 and K2 are thermal conductivity, V1 and V2 are the volumefractions of phase 1 and 2, respectively. Fig. 7 compares the exper-imental values with those evaluated using parallel model. The mea-sured values apparently agree very well with the predictions exceptfor composites with over 80 vol.% of graphite. The larger deviationbetween the measured and predicted thermal conductivity valuesfor composite containing 90 vol.% is apparently caused by its sinter-ing density below 93% which reduces the graphite connections.

3.3. Thermal expansion of Al–graphite composites

The coefficients of thermal expansion of aluminum–graphitecomposites are measured along two directions, namely parallel tographitebasalplane(a-axis)andperpendiculartobasalplaneoralongc-axis. Three models are employed to compare with the measuredcoefficients of thermal expansion for current composite system:

(a) Rule of mixture

aM ¼ V1a1 þ V2a2 ð2Þ

(b) Turner model [24]:

aM ¼a1B1V1 þ a2B2V2

B1V1 þ B2V2ð3Þ

(c) Kerner model [25]:

aM ¼ a1 þ V2ða1 � a2Þ

� B1ð3B2 þ 4G1Þ2 þ ðB2 � B1Þð16G21 þ 12G1B2Þ

ð4G1 þ 3B2Þ½4V2G1ðB1 � B2Þ þ 3B1B2 þ 4G1B1�ð4Þ

Fig. 7. Comparison of measured and parallel model predicted thermal conductivity.

where a stands for coefficients of thermal expansion, V is the vol-ume fraction, B is the bulk modulus, G is shear modulus, and thesubscripts M, 1, and 2 refer to the composite, matrix and fillerphase, respectively.

Fig. 8 shows the measured and estimated coefficients of thermalexpansion (CTE) results. The CTE of graphite along a-axis is�1.5 ppm/K and that along c-axis is 8 ppm/K [26,27], and CTE ofaluminum is 23.8 ppm/K. The measured CTE both along and per-pendicular to basal plane direction of aluminum–graphite compos-ites decreases with graphite but with very different trends.

The CTEs of composites in directions parallel to the basal planeof graphite agree reasonably well with rule of mixture as shown inFig. 8a. According to the rule of mixture, thermal expansion ofcomposites constitutes of proportional contributions from bothaluminum and graphite according to their volume percentages.Therefore, the good approximations of experimental values andrule of mixture predictions are representative of good interfacebonding between the two phases. The effects of mechanical prop-erties, such as bulk modulus and shear modulus, affect not muchon the CTEs of the composites since Kerner model does not fit wellwith the measured CTEs. Because aluminum and graphite flakesare both soft in nature, the two phases can expand freely alongthe directions parallel to the basal planes. Since the expansion ofaluminum and graphite is additive by proportion, the shear stressat the aluminum–graphite interfaces must also be minimal with-out restraining either phase to expand.

On the other hand, the CTEs of composites along c-axis or per-pendicular to basal planes are more consistent with Turner modelestimations. Turner model [24] is developed for predicting com-posites constituted by phases with similar modulus and is based

Fig. 8. Experimental and theoretical coefficients of thermal expansion for compos-ites containing different percentage of graphite along (a) a-axis and (b) c-axisdirection.

702 J.K. Chen, I.S. Huang / Composites: Part B 44 (2013) 698–703

upon zero internal stress assumption. Because graphite flake alongc-axis direction is characteristic of Van der Waal bonds, the modu-lus along this direction is fairly low and is comparable with that ofpure aluminum. In Fig. 8b, the CTEs decrease rapidly with increas-ing graphite content to �15 ppm/K and keep at a relatively con-stant level close to the CTE of graphite in c-direction. It indicatesthat the expansion of aluminum is mostly absorbed by graphiteflake in direction perpendicular to basal planes. The expansion ofaluminum becomes omittable for aluminum–graphite compositescontaining graphite of nearly all percentage. This still holds trueeven for composites with high porosity at very high graphite-containing composites.

Graphite networks in direction perpendicular to basal planesappear to absorb the expansion of the binder phase. The poresamong graphite flake networks and weak Van der Waal’s force ingraphite flakes along c-axis can facilitate the absorption.

4. Conclusions

Vacuum hot press in semi-solid liquid state is employed tomake aluminum–graphite composites. Over 98% densification isachieved for composites containing lower than 70 vol.% of graph-ite. The interface is observed to bear good bonding characteristicswhile no detrimental Al4C3 phase is found to form in composite.Aluminum acts as the bonding phase between graphite flakes.Porosity increased to over 5 vol.% when flake graphite content isincreased over 80 vol.% due to lack of aluminum diffusion intographite networks. The pores are observed to exist at the interfacesbetween the graphite flakes.

Thermal conductivity of aluminum–graphite compositesincreases with increasing amount of graphite. For composites

containing 10–80 vol.% of graphite, the measured thermal conduc-tivity increases from 324 to 783 W/m K. It agrees well with theparallel model indicating the good interfaces in between alumi-num and graphite. Larger amount of pores in aluminum–90 vol.%graphite composites causes its thermal conductivity to deviate to-ward lower values.

Coefficients of thermal expansion for aluminum–graphite com-posites decrease with increasing flake graphite. Good bondinginterfaces and low mechanical strengths of aluminum and graphitecause the coefficients of thermal expansion along the basal plane todemonstrate a trend close to rule of mixture. The values range from16.9 to �2.5 ppm/K with increasing graphite content. On the otherhand, the coefficients of thermal expansion in direction perpendic-ular to the basal planes of graphite in composites demonstratecomparably higher values mainly due to the low Van der Waal forcealong this direction. The measured values fits well with Turnermodel indicating that the component of softer graphite directionprovides an absorber for the expansion of aluminum in directionperpendicular to the basal planes. Therefore the CTE is closer to thatof graphite along c-direction.

The current study demonstrates that the thermal conductivityand thermal expansion of aluminum–graphite composites can becontrolled easily by the percentage of graphite and orientationvia hot pressing. The process is shown to achieve excellent inter-face bonding and can avoid detrimental interfacial reactionsbetween aluminum and flake graphite.

Acknowledgment

The authors are grateful for financial support of this project byNational Science Council of Taiwan under Contract #NSC99-2221-E-027-020.

References

[1] Luedtke A. Thermal management materials for high-performance applications.Adv Eng Mater 2004;6(3):142–4.

[2] Zweben C. Advances in composite materials for thermal managementmaterials in electronic applications. JOM 1998;50(6):47–51.

[3] Li DS, Garmestani H, Schwartz J. Modeling thermal conductivity in UO2 withBeO additions as a function of microstructure. J Nucl Mater 2009;392(1):22–7.

[4] Rawdanowicz TA, Godbole V, Narayan J, Sankar J, Sharma A. Hardnesses andelastic moduli of pulsed laser deposited multilayer AlN/TiN thin films. ComposPart B: Eng 1999;30(7):657–65.

[5] Chen AL, Arai Y, Tsuchida E. An experimental study on effect of thermal cyclingon monotonic and cyclic response of cast aluminum alloy-SiC particulatecomposites. Compos Part B: Eng 2005;36(4):319–30.

[6] Lorentzen T, Clarke AP, Poulsen HF, Garbe S, Graafsma H. Local strain contoursaround inclusions in wire-drawn Cu/W composites. Compos Part A: Appl SciManuf 1997;28(7):667–74.

[7] Zweben C. Advanced electronic packaging materials. Adv Mater Proc2005;163(10):33–7.

[8] Kawai C. Effect of interfacial reaction on the thermal conductivity of Al–SiCcomposites with SiC dispersions. J Am Ceram Soc 2001;84(4):896–8.

[9] Lambert MA, Fletcher LS. Thermal conductivity of graphite/aluminum andgraphite/copper composites. J Heat Trans 1996;118(2):478–80.

[10] Arpon R, Molina JM, Saravanan RA, Garcia-Cordovilla C, Louis E, Narciso J.Thermal expansion behaviour of aluminum/SiC composites with bimodalparticle distributions. Acta Mater 2003;51:3145–56.

[11] Pal S, Bhanuprasad VV, Mitra R. Damage evaluation in discontinuouslyreinforced aluminum matrix composites subjected to thermal cycling. MaterSci Eng A 2008;489(1–2):11–20.

[12] Yang J, Chung DDL. Casting particulate and fibrous metal–matrix compositesby vacuum infiltration of a liquid metal under an inert gas pressure. J Mater Sci1989;24(10):3605–12.

[13] Lai SW, Chung DDL. Fabrication of particulate aluminium–matrix compositesby liquid metal infiltration. J Mater Sci 1994;29(12):3128–50.

[14] Etter T, Kuebler J, Frey T, Schulz P, Loffler JF, Uggowitzer PJ. Strength andfracture toughness of interpenetrating graphite/aluminium compositesproduced by the indirect squeeze casting process. Mater Sci Eng A2004;386(1–2):61–7.

[15] Etter T, Schulz P, Weber M, Metz J, Wimmler M, Loffler JF, et al. Aluminumcarbide formation in interpenetrating graphite/aluminum composites. MaterSci Eng A 2007;448(1–2):1–6.

J.K. Chen, I.S. Huang / Composites: Part B 44 (2013) 698–703 703

[16] Ueno T, Yoshioka T, Sato K, Makino Y, Endo M. Fabrication of high thermalconductive aluminum/graphite fiber composites by pulsed electric currentsintering. J Jpn Soc Powder Powder Metall 2007;54(8):595–600.

[17] Mizuuchi K, Inoue K, Agari Y, Morisada Y, Sugioka M, Tanaka M, et al.Processing of diamond particle dispersed aluminum matrix composites incontinuous solid–liquid co-existent state by SPS and their thermal properties.Compos Part B: Eng 2011;42(4):825–31.

[18] Maire J, Mering J. Graphitization of soft carbons. In: Walker PL, editor.Chemistry and physics of carbon, vol. 6. New York: Marcel Dekker; 1970.

[19] Jiang N, Novak JP, Fink RL. Fabrication and characterization of carbon–aluminum thermal management composites. In: Proceedings of 26th annualIEEE semiconductor thermal measurement and management symposium,SEMI-THERM 2010. Santa Clara, CA, February, 2010. p. 87–92.

[20] Chang CJ, Chang CH, Hwang JD, Kuo CT. Thermal characterization of highthermal conductive graphites reinforced aluminum matrix composites. In:Proceedings of IMPACT conference 2009 international 3D IC conference. Taipei,Taiwan, October, 2009.

[21] Baglari S, Kole M, Dey TK. Effective thermal conductivity and coefficient oflinear thermal expansion of high-density polyethylene-fly ash composites. IndJ Phys 2011;85(4):559–73.

[22] Yoshida K, Morigami H. Thermal properties of diamond/copper compositematerial. Microelectron Reliab 2004;44(2):303–8.

[23] Truong HV, Zinsmeister GE. Experimental study of heat transfer in layeredcomposites. Int J Heat Mass Trans 1978;21(7):905–9.

[24] Turner PS. Thermal-expansion stresses in reinforced plastics. J Res NBS1946;37:239–50.

[25] Kerner EH. The elastic and thermo-elastic properties of composite media. ProcPhys Soc Sec B 1956;69(8):808–13.

[26] Nelson JB, Riley DP. The thermal expansion of graphite from 15 �C to 800 �C:Part I. Experimental. Proc Phys Soc 1945;57(6):477–86.

[27] Riley DP. The thermal expansion of graphite: Part II. Theoretical. Proc Phys Soc1945;57(6):486–95.