Microstructure and thermophysical properties of graphite foam/Sn-Bi alloy composites for use as a thermal sink for electronics

LIAN Peng-fei, SONG Jin-liang, LEI Shi-wen, TAO Ze-chao, ZHAO Hong-chao, ZHANG Jun-peng, LIU Zhan-jun

(1. Key Laboratory of Carbon Materials, Institute of Coal Chemistry, Chinese Academy of Sciences, Taiyuan030001, China; 2. Key Laboratory of Nuclear Radiation and Nuclear Energy Technology, Shanghai Institute of Applied Physics, Chinese Academy of Sciences, Shanghai201800, China)

Abstract: Two mesophase pitch-based graphite foams with densities of 0.62±0.01 (GF1) and 0.84±0.01 g/cm3 (GF2) were prepared by foaming the pitch in an autoclave at 723 K, 6.0 MPa and 763 K, 13.4 MPa, respectively, followed by carbonization at 1273 K for 2 h and graphitization at 2973 K for 0.5 h. The GFs were infiltrated by a Sn-Bi liquid to prepare GF/Sn-Bi alloy composites for use as thermal sinks for electronics. The microstructures and thermophysical properties of the composites were investigated. Results indicated that GF1 had a larger cells and thinner cell walls than GF2. The Sn-Bi liquid was well infiltrated into cells of the GFs, resulting in composites with densities of 5.60±0.01 and 3.83±0.01 g/cm3 for GF1 and GF2, respectively. The thermal diffusivity and coefficient of thermal expansion (CTE) of the GF1/Sn-Bi composite were 51.6±2 mm2/s and 16.6±0.02 ppm/K, respectively. The corresponding values for the GF2/Sn-Bi were 163.1±3 mm2/s and 8.08±0.02 ppm/K. The GF2/Sn-Bi composite had a high thermal diffusivity and a low CTE value matching that of semiconductor chips and packaging materials.

Key words: Mesophase pitch-based graphite foam; Sn-Bi alloy; High thermal conductivity; Low coefficient of thermal expansion

1 Introduction

Thermal management and thermal stresses are critical issues in many electronic devices, including microprocessors, power semiconductors, high-power radio frequency devices, light-emitting diodes[1-3]. Although thermal management is a complex issue, it is clear that packaging materials with a high thermal diffusivity and tailorable coefficient of thermal expansion (CTE) ranging 3-10 ppm/K are needed to ensure the reliability and performance of semiconductor chips[1-4]. However, with the increase of power density, the properties of the conventional electronic packaging materials cannot well satisfy the demands of electronic devices[1,4]. Mesophase pitch (MP)-derived graphite foams are composed of inherent interconnected networks of ligaments or struts, which make them suitable candidates as the matrix of high-thermal-conductivity composites[5-7]. Furthermore, the foams exhibit an isotropic feature, leading to the formation of isotropic graphite foam composites that are different from the anisotropic type continuous carbon fiber composites[8]. Klett et al.[5]suggested that the ligaments and cell walls of the foams exhibit thermal conductivites greater than 1 500 W/(m·K) based on the analysis of X-ray diffraction. This suggests that the thermal conductivity of graphite foam composites may be much higher if the foam framework is not damaged or the foam possesses a higher density[6]. Therefore, some composites involving graphite foams have been developed for thermal management. Klett et al.[6]investigated the thermal properties of graphite foams (0.54 g/cm3) infiltrated with several polymers. The thermal conductivity of the graphite foam/polycyanate composites reaches 129 W/(m·K) at room temperature. McCoy and Vrable[8]reported that graphite foam/copper composites exhibited a low CTE of 7.43 ppm/K and a high thermal conductivity of 342 W/mK. A pressureless infiltration method has been used to prepare graphite foam/copper composites which is realized by the synthesis of Mo2C coating on the graphite foam walls. Thermal conductivity of the obtained graphite foam/copper composites reaches 268.4 W/(m·K) and the average CTE is decreased to 8.91 from 18.59 ppm/K of copper. However, the high infiltration temperature and cost limit the wide application of the composites as electronic package materials[9].

Sn/Bi alloy is an important welding and electronic packaging material. The alloy matrix is attractive due to its tailorable thermal conductivity and CTE, compared with most of the pure metals and polymers[2,10]. Thus, the thermophysical properties of the composites can be retained during manufacturing and thermal cycling. Furthermore, the infiltration temperature of alloy into graphite foams can be controlled by changing its composition[10]. Unfortunately, to the best of our knowledge, the fabrication technique of graphite foam/alloy composites and their thermophysical properties have not been focused.

In the present work, two graphite foam/Sn-Bi alloy composites have been developed at the infiltration pressure of 2 MPa. Two important performance parameters of electronic packaging materials, namely thermal diffusivity (thermal conductivity) and CTE, have been characterized. The effects of the structure and thermophysical properties of graphite foams on the thermal behavior of the composites have been discussed.

2 Experimental

2.1 Sample preparation

Mesophase pitch synthesized from Mitsubishi naphthalene was used to produce graphite foams in an autoclave. Two kinds of graphite foams were prepared and denoted as GF1 (with a density of 0.62±0.01 g/cm3)and GF2 (with a density of 0.84±0.01 g/cm3). The final foaming temperatures and pressures for the GF1 and GF2 were 723 K, 6.0 MPa and 763 K, 13.4 MPa, respectively. The foams were carbonized at a heating rate of 0.5 K/min to 1 273 K, held for 2 h and then graphitized at 2 973 K for 0.5 h. One kind of alloy material (a weight ratio of Sn/Bi = 4∶1, a melting point of ～483 K) was infiltrated into the two graphite foams at 493 K under 2 MPa and kept under 2 MPa for 1 h. The final composites were denoted as GF1/alloy and GF2/alloy, respectively. The basic properties of the graphite foams and alloy were listed in Table 1.

Table 1 Properties of the graphite foams and alloy.

2.2 Sample characterization

Scanning electron microscopic (SEM) images obtained on a LEO 1530VP were used to characterize the average pore diameters and microstructures of the samples. The density of the samples was determined by a densitometer (Accupy 1 340 Micromericics USA). Thermal diffusivity (a) of the samples (Sample size: 10 mm × 10 mm × 4 mm) were measured by a Netzsch LFA447/2-2 InSb Nano Flash machine at room temperature. Thermal conductivity (k) was calculated by the formulak= a · Cp·ρ, whereCpandρwere the specific heat capacity and bulk density of each sample, respectively. TheCpof the alloy was characterized by the differential scanning calorimeter technique (NETZSCH DSC 204F1). TheCpof the composite was calculated based on the “rule of mixtures” and the skeleton density of graphite foam was 2.25 g/cm3[8,11-13]. The CTEs were tested by a dilatometer (DIL 402 PC NETZSCH. Sample size: 3.5 mm × 3.5 mm × 25 mm). Temperature was varied from 298 to 373 K with a heating rate of 3 K/min.

3 Results and discussion

3.1 Microstructure

Fig. 1a shows the XRD patterns of the Sn-Bi alloy. The peaks at 30.6°, 32.0°, and 44.9° indicated the characteristic peaks of Sn. While the peaks at 27.2°, 39.6°, and 37.9° were the characteristic peaks of Bi. Fig. 1b shows the specific heat capacity of the Sn-Bi alloy. The specific heat capacity at room temperature was 0.11 J/gK and the peak temperatures of phase transformation appeared at 419 K and 485 K. The phase change of the alloy implied the volume change at the two temperatures, which might lead to fractures during the preparation of the Sn-Bi alloy and packaging matrix. Enhancing the thermal conductivity could limit the expansion of the Sn-Bi alloy, ensure the quick heat dissipation and dimensional stability, and inhibit the fracture of the package materials.

Fig. 1 (a) XRD patterns, (b) the specific heat capacity of the Sn-Bi alloy and (c, d) the EDS line-scan analysis across the interface of GF and alloy.

Fig. 1c and d show EDS line-scan analysis of the major element distribution in the interface zone between the GF and Sn-Bi alloy. The interpenetrated interface structures and interconnected graphite network were expected to reinforce the foam matrix and to constrain the CTE of alloy effectively by utilizing the mechanical interlocking and the space limitations of the graphite network.

Fig. 2a and b show the microstructures of GF1 and GF2. Both foams exhibited a spherical structure with open, interconnected pores (P in Fig. 2a) and microcracks in graphite walls (W in Fig. 2a). The interconnected structure facilitated the infiltration of molten alloy into graphite foams. It can be seen that there were significant differences in the cell sizes and microstructures between GF1 and GF2. GF1 possessed an average cell size of ～ 300 μm and many big pores of ～ 100 μm between the cells at the connection of two big pores. GF2 had thicker foam walls, a smaller average cell size of ～ 150 μm and fewer big pores. Fig. 2c and d show the cross sectional SEM images of the GF1/alloy and GF2/alloy samples, respectively. The foams had a good contact with alloy and a high degree of alloy infiltration was realized, even at the smallest cell size of ～ 100 μm.

Fig. 2 SEM images of (a) GF1 (P-cell pore, W-foam wall), (b) GF2, (c, e) GF1/alloy (F- flat faces) and (d, f) GF2/alloy.

Fig. 2e and f show the fracture surfaces of GF1/alloy and GF2/alloy, respectively. The nearly spherical alloy particles were embedded within the foam cells. Some of the alloy particles were broken and the flat faces (F in Fig. 2d and e) had been formed during the fracture of the composites. The foams and alloy combined compactly without obvious interfacial separation or cracks in the composites. The alloy particles in GF1 were mainly connected with each other by big pores in the foam. Due to the existence of fractures across these big pores at the joint of the two cells, some alloy particles had flat faces (F in Fig. 2e). The molten alloy was infiltrated into GF2 mainly through the microcracks in the foam walls and the alloys could not be fully infiltrated in the foams, leading to a low density of 3.83±0.01 g/cm3of the GF2 composites. Before the alloy particles were pulled out from the cell, the alloy particles were surrounded by the graphite flakes of the foam (Fig. 2c and d). It could be observed that the alloy particles were embedded in the foam cells. After the alloy particles were pulled out from the cells, the fractograph of the two particles at the big pore between the two cells is shown in Fig. 2c. The ligaments and cell wall of the cell framework were retained well after alloy infiltration. All the microcracks in the cell wall were saturated with alloy (Fig. 2c), which was crucial for the complete infiltration of alloy into the GF1 to attain the high density of 5.60±0.01 g/cm3of the composites. The smooth cell wall without obvious fractures (Fig. 2e and f) indicated poor chemical interfacial bonding between the graphite foam and the alloy, which was favorable to minimize thermal stress during the casting process and thermal cycling. The graphite foam was considered as rigid network, while the alloy particles as the elastic balls were connected by elastic alloy rods. During solidification and thermal cycling, the shrinkage and expansion of the two phases were nonsynchronous. Strong interfacial bonding and compact structure might damage the foam framework, leading to the degradation of thermal properties of the foams[6]. In addition, the mechanical interlocking and space limitations of the foam network could play a crucial role in limiting the thermal expansion of the alloy, which will be discussed in the later section.

3.2 Thermophysical property

Table 2 lists the properties of the Sn-Bi alloy and graphite foam/alloy composites. The thermal diffusivity of GF1/alloy and GF2/alloy reached 51.6±2 and 163.1±3 mm2/s, which were 2.0 and 6.4 times of the Sn-Bi alloy, respectively. The heat capacity and bulk density of GF2/alloy were 0.40 J/gK and 3.83±0.01 g/cm3, respectively. The thermal conductivity of GF2/alloy reached 249.9±4.6 W/mK, which was higher than that of aluminum and most of the electronic substrates and packaging materials[1-4, 10]. The calculated results based on the mass-density relationship indicated that the alloy filled more than 93% and 64% of the available pore volume of GF1 and GF2, respectively. The alloy was uniformly dispersed in the graphite cells or into the cell walls (Fig. 2c), which increased obviously the total contact area between the graphite foam and the alloy. Excellent continuity at the interfaces and throughout the foam supplied good pathways for heat transfer. The high thermal conductivity of the ligaments and the walls in graphite foams ensured a rapid heat transfer throughout the composites. The smaller the pores (P in Fig. 2a) and average cell size of the foams, the larger the contact area between graphite foams and the alloy, and the higher thermal diffusivities of the composites. As for the GF2/alloy composite, the average cell size was small and most alloy particles were connected with each other by the alloy saturating in the foam microcracks, which led to more effective heat dissipation from the composites.

Table 2 Properties of the graphite foam/alloy composites and copper.

Fig. 3 show the thermal expansion curves of GF1, GF2, alloy, GF1/alloy and GF2/alloy after ten thermal cycles from 298 to 373 K. The average CTEs of graphite foams and the corresponding composites are listed in Table 1 and Table 2, respectively. As can be seen, the GF2 with a CTE of 1.71±0.02 ppm/K could decrease the CTE of the Sn-Bi alloy to 8.08±0.02 ppm/K. The alloy based composites had been subjected to ten thermal cycles from 298 to 373 K at the heating and cooling rates of 3 K/min. The thermal expansion curves of the composites mutually entangled. No obvious CTE changes could be identified before and after the thermal cycles, which indicated that the composites possessed an excellent thermal stability. When the composites were heated, the faster expansion of the alloy was constrained by the graphite network. This became a complex tug-of-war where the strain mismatch was resolved into stresses at the interfaces where the two interpenetrated phases met. The mechanical interlocking and the space limitations of the foam network plays a crucial role on constraining the thermal expansion of the alloy. The space limitations caused by the high modulus of the graphite contributed much greater to constraining the thermal expansion. Owing to the thermal stability of GF2 composite, it is expected as the promising candidate of electronic packaging materials used in military, aviation, aerospace, et al. The average CTE of GF2/alloy was 8.08±0.02 ppm/K, which was in good agreement with that of AlN substrate (4.3 ppm/K) and semiconductor chips such as Si (4.1 ppm/K), GaAs (5.8 ppm/K) and InP (4.8 ppm/K)[1-4]. The CTEs of the composites could be adjusted by changing the volume fractions of graphite foams in the composites. As to the two phase composite, one may expect the CTE of the composite to follow a simple "rule of mixtures" for composites[14]given by the equation:

αc=αaVa+αfVf

(1)

Whereαis the CTE value,Vis the volume fraction, and subscripts c, a, f refer to the composite, alloy and graphite foam, respectively. As shown in Table 2, the experimental CTE value of GF2/alloy was smaller than the calculated ones according to Eq.(1). It was in good agreement with the reported results when interconnected materials were used as effective reinforcements[8,11,14]. The foam reinforced composites possessed lower CTEs than the simple two phase composite like particle/fiber reinforced composites due to the space limitation of the foam skeleton structure[14]. For the SiC/Al composites, SiC foam reinforced composite also exhibited a lower CTE than that of SiC particle reinforced one with the same volume fraction of SiC[14].

Fig. 3 Thermal expansion curves of GF1, GF2, alloy, GF1/alloy and GF2/alloy after ten thermal cycles from 298 to 373 K.

4 Conclusions

Graphite foam/Sn-Bi alloy composites with high thermal diffusivities and low CTEs were prepared. Microstructure analysis indicated that the alloy was well dispersed in the graphite cells or into the cell walls. The thermal diffusivity of the composite with a density of 3.83±0.01 g/cm3reached 163.1±3 mm2/s, 6.4 times that of the alloy. The CTEs of the composites decreased from 16.40±0.02 to 8.08±0.02 ppm/K, and the CTE of the high density foam (GF2/alloy) composite decreased from 20.70 of the alloy to 8.08±0.02 ppm/K of the obtained composite, which matched the CTE requirements of semiconductor chips and packaging materials well. The composites exhibited an excellent dimensional stability. After ten thermal circles between 298 and 373 K, the CTE values of these two composites were essentially maintained.