Materials and Design 30 (2009) 1098–1102

Contents lists available at ScienceDirect

Materials and Design journal homepage: www.elsevier.com/locate/matdes

Roles of preoxidation, Cu2O particles, and interface pores on the strength of eutectically bonded Cu/a-Al2O3 H. Ghasemi a,*, A.H. Kokabi a, M.A. Faghihi Sani a, Z. Riazi b a b

Department of Materials Science and Engineering, Sharif University of Technology, Tehran, Iran Bonab Research Center, Bonab, Iran

a r t i c l e

i n f o

Article history: Received 27 February 2007 Accepted 17 June 2008 Available online 24 June 2008 Keywords: Ceramic–metal (A) Bonding (D) Mechanical strength (E)

a b s t r a c t The influences of CuO layer thickness, Cu2O particles, and pores on mechanical properties and microstructure of alumina–copper eutectic bond have been investigated. The furnace atmosphere in the first stage was argon gas with 2  106 atm oxygen partial pressure. In the second stage, the furnace atmosphere was same as the first stage except for the cooling interval between 900 and 1000 °C, the hydrogen gas was injected into furnace atmosphere. Finally, in the last stage a vacuum furnace with 5  108 atm pressure was chosen for bonding procedure. Peel strength of first stage specimens shows that CuO layer with 320 ± 25 nm thick generates the maximum peel strength (13.1 ± 0.3 kg/cm) in joint interface. In the second stage, by using the hydrogen gas, a bond interface free of any Cu2O oxide particle was formed. In this case, the joint strength has increased to 17.1 ± 0.2 kg/cm. Finally, the bonding process in vacuum furnace indicates that the furnace gas does not have considerable effect on joint interface pores. Furthermore, bonding process in vacuum furnace reduces the peel strength of joint due to formation of more pores. Thorough study of pores formation is presented. Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction Gas–metal eutectic bonding of alumina–copper was first developed by Burgess and Neugebauer [1]. So far, the mechanical properties of eutectic bonding and corresponding interactions within the bonding area have been studied extensively by several authors [2–8]. The bonding of copper to alumina is carried out at 1075 ± 2 °C in a properly controlled oxygen-containing atmosphere. It has been demonstrated that oxygen promotes wetting of liquid copper to alumina and hence it is necessary to achieve good spreading of the liquid along the bond interface [2,9]. Yoshino [2] observed that the peel strength of alumina–copper bond is predominantly affected by dissolved-oxygen concentration. He reported that the Cu2O particles and voids have, also, contributions to true bond strength. Thus, oxygen seems to affect the peel strength of alumina/copper bond principally in two ways: through change in cohesive bond energy and through formation of Cu2O particles. The measured bond strength is the combination of these effects. Yoshino predicted that by removing the oxide particles from the interface, the peel strength of the joint can be raised up to 15 kg/cm [2]. Moreover, Trumble et al. [7] have reported that the Cu2O particles can be reduced after bonding without any loss in bond strength. Kim and Kim [5] studies on sandwich type Cu– * Corresponding author. Tel.: +1 416 543 9757. E-mail address: [email protected] (H. Ghasemi). 0261-3069/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.matdes.2008.06.034

Al2O3 couple shows that by creating CuAlO2 phase at the interface, the bond strength increases up to an optimum CuAlO2 layer thickness which after that brittleness of CuAlO2 lowers the bond strength. Likewise, Seager et al. [6] have investigated more precisely the effect of CuAlO2 layer thickness and have confirmed that thick CuAlO2 layer is detrimental to bond strength. Following the other researcher investigations, Ning et al. [8] have stated that good bonding strength can be obtained by a certain thickness of CuO layer which is a function of alumina substrate profile. On source of pores formation and their effect on peel strength, Yoshino [2] reported that the pores within the bonding interface can be formed due to release of oxygen gas in the liquid copper; by increasing the CuO layer thickness, their percentage and size will be larger. However, Seager et al. [6] have found that smaller pores (1–3 lm) may be pullouts of the Cu2O particles observed on alumina fracture surface, while the large pores observed by Yoshino are the result of argon entrapment in liquid copper during processing. Concerning the influence of pores on mechanical strength of eutectic bond, Reimanis [10] have observed that crack-front perturbation occurs when the crack tip is in contact with a pore; the crack front is drawn into the pore and causes debonding of the regions immediately surrounding the pore. Despite several investigations, the accurate contribution of preoxidation and Cu2O particles on peel strength of mentioned joint is unclear. There is, furthermore, uncertainness about the sources of pores and their effect on peel strength of the bond. The goal of current research is to shed

H. Ghasemi et al. / Materials and Design 30 (2009) 1098–1102

more light on the effects of CuO layer thickness, Cu2O particles, and pores on peel strength of alumina–copper eutectic bond. Besides, the sources of pores formation have investigated thoroughly. 2. Experimental procedures and apparatus Experiments were carried out in three stages. In the first stage, the effect of preoxidation on peel strength of alumina–copper bond has been examined. In the second stage, influence of Cu2O particles on peel strength of alumina–copper bond has been investigated. Finally, in order to decrease the interface pores, the bonding process was performed in a vacuum furnace. All of the alumina–copper bonds began with 99.99% Cu strips and 97% Al2O3. Copper strips and alumina specimens were 350 and 200 mm in length, 25 mm in width, and 0.8 and 1.5 mm thick, respectively. Copper strips were prepared in three steps. Firstly, strips were degreased by scrubbing the joint surfaces in a solution of liquid detergent, washing with clean hot water and drying thoroughly in a steam of hot air. Secondly, strips were immersed in a 25% nitric acid for 30 s. Thirdly, strips were dipped in ethyl alcohol followed by clean cold water and drying by hot air. On the other hand, alumina specimens were polished by diamond paste with average particle size of 0.5 lm, cleaned by ultrasonic treatment in ethyl alcohol for 15 min and rinsed in distilled water. At the final step, alumina specimens were annealed in air at 1000 °C to eliminate any hydroxyl groups. Copper strips were preoxidized in various temperatures and times to obtain a range of CuO layer thickness. On account of phase stability diagram of copper oxidation, formed oxide phase below 400 °C under atmosphere pressure will be CuO. The thickness of CuO layer was calculated from the weight difference before and after preoxidation. In the first stage, the sample were bonded in a tube furnace in argon atmosphere (<2 ppm O2). Gas flow in heating and cooling were 150 and 200 L/h, respectively. The samples were heated at a rate of 10 °C/min to 1000 °C, followed by rate of 2 °C/min to 1075 °C where they were held for 1 h. During the holding time, alumina–copper bond was formed. After 1 h, the temperature decreased slowly (5 °C/min) to 700 °C and then 10 °C/min to 400 °C. The temperature was held for 1 h in 400 °C to reduce thermal stresses. Finally, the furnace cooled to room temperature. The furnace temperature curve is illustrated in Fig. 1. In the second stage, bonding condition was as same as first stage in expect that in cooling interval between 900 and 1000 °C the hydrogen gas was injected into the furnace atmosphere. The purity of hydrogen gas and its flow rate were 99.99% (Thc <0.2) and 100 L/h, respectively. In the last stage, bonding process was carried out in a vacuum furnace. The vacuum tube with a mechanical pump and an oil diffusion pump can make an approximately vacuum atmo-

Fig. 1. The furnace curve in bonding process.

1099

sphere of 5  108 atm. The oxygen partial pressure of oxygen was measured by CaO-stabilized ZrO2 oxygen sensor. The peel strength of specimens was measured by Instron Machine Model 1115. The fracture surfaces were analyzed by optical microscopy and scanning electron microscopy. 3. Results and discussion 3.1. CuO layer thickness The thickness of CuO layer formed in various temperatures and times is depicted in Fig. 2. It shows that the thickness variation of the CuO layer is gradual and uniform within the range of 0–600 nm above which the thickness increase is abrupt and discontinuous. Samples having an oxide layer thickness greater than 600 nm present a discontinuous and cracked oxide layer surface. Furthermore, the adhesion between the CuO layer and copper strips has decreased notably. These discontinuities on oxide layer surface are due to thickness of CuO layer. The more oxide layer thickness, the more sensitivity to thermal expansion coefficient mismatch. In as much as the expansion coefficient of Cu (17.6  106 °C1) and CuO (4.3  106 °C1) have great difference, the formed residual stresses in cooling is high. According to modified Stoney equation (Eq. (1)) [11], the formed normal stress at the interface can be up to 2 Gpa.



R Td Eef Tr ðas  af ÞdT 1 þ 4ðEef =Ees Þðh=HÞ

ð1Þ

where Eef [=Ef/(1  mf)], Ees [=Es/(1  ms)], ms, mf, h, H, Td, Tr, as, and af are effective Young’s modulus of the coating, effective Young’s modulus of the substrate, Poisson ratio of the substrate, Poisson ratio of the coating, coating thickness, substrate thickness, deposition temperature, room temperature, thermal expansion coefficients of the substrate and the coating, respectively. Actually, since the temperature difference between substrate and coating is lower than 400 °C (maximum oxidation temperature), the formed stresses will be lower than this amount. However, they are sufficient to break CuO/Cu bond. The thermal stress field generates discontinuities and cracks within the CuO layer which in turn results in an increase of the copper oxidation rate. Higher oxidation rate is due to greater copper exposure to oxidizing atmosphere and active oxidation. This fact is obvious in Fig. 2. Hence, the preoxidation conditions which create more than

Fig. 2. The variation of oxide layer thickness in different preoxidation conditions.

1100

H. Ghasemi et al. / Materials and Design 30 (2009) 1098–1102

approximately 500–600 nm oxide layer thickness, makes the strips improper for bonding process. 3.2. Peel strength measurements The peel strengths of first stage specimens are shown in Fig. 3. As shown in Fig. 3a, an optimum oxide layer thickness results in

Fig. 3. The peel strength of specimens versus CuO layer thickness.

the highest peel strength (13.1 ± 0.3 kg/cm) under a ‘‘2  106’’ atm oxygen partial pressure. This optimum thickness is a result of the interaction between wetting of alumina surface and percentage of Cu2O particles at the interface. Thicker CuO layer introduce more oxygen in interface and subsequently the wetting angle will be decreased [11] and more Cu2O particles will remain in interface after bonding. Similarly, Fig. 3b shows an optimum point. By comparison the optimum layer thickness in Fig. 3a and 3b, there is approximately 25 nm difference. This difference can be resulted from errors in peel strength and oxide layer thickness measurement. Thus, it can be said that 320 ± 25 nm is the optimum oxide layer thickness for alumina–copper bonding in 2  106 atm oxygen partial pressure. The peel strength difference in these two states is low; it shows that oxidation in different temperatures to obtain certain oxide layer thickness does not affect the peel strength. Both Fig. 3a and 3b reveal that before reaching optimum point, the peel strength of specimens has raised steadily. It can be concluded that improved wetting due to increase in the oxygen content of liquid copper has heighten the peel strength. After optimum point, the effect of Cu2O content is predominant and has decreased the bond strength. Fig. 3c shows the peel strength of specimens oxidized in 400 °C. In contrast to Fig. 3a and 3b, in preoxidation time span of this temperature, the peel strength of specimens has decreased slowly. Obviously, in this time span, the Cu2O content of interface layer is predominant parameter to wetting of alumina by liquid copper. The noteworthy point in Fig. 3a and 3b is the rising and falling rate of peel strength versus the oxide layer thickness. It shows that the effect of wetting on peel strength is more significant than effect of Cu2O content of interface. This state is, also, clear in Ning investigations [8]. To achieve more strength in alumina–copper bond, it is desirable to remove Cu2O from the interface. This fact is due to higher strength of Cu/Al2O3 bond comparing with Cu/Cu2O and Cu2O/ Al2O3 [2,4,7]. Hence, during the cooling interval of 900–1000 °C within the second stage, the hydrogen gas was injected into the furnace atmosphere in order to remove Cu2O particles from the eutectic zone of the interface. Fig. 4 shows the cross-section of alumina–copper joint in the second stage. It, surprisingly, illustrates no trace of interface layer; the joint is directly established between alumina and copper. No sign of any other phase in expect of copper and alumina is detectable at the fracture surfaces. The peel strength measurements of the joint in this stage are shown in Fig. 5. In addition, Fig. 6 depicts the force versus distance curve of maximum peel strength in first and second stage. As shown in Fig. 5, in contrast to first stage specimens, the peel strength of the joint has raised considerably in all CuO layer thicknesses.

Fig. 4. The cross-section of alumina–copper bond after removal of the Cu2O particles.

H. Ghasemi et al. / Materials and Design 30 (2009) 1098–1102

1101

Fig. 5, is due to an increase of the pore density on the bonding interface. The change of CuO optimum layer thickness in this stage was predicted. Due to reduction of Cu2O particles, the effect of wetting is more significant than an increase of pore0 s concentration and distribution for increasing oxygen content of liquid copper; resulting in a shift of the optimum peel strength to the values corresponding to a thicker CuO layer. By reducing one mole of Cu2O, two mole of Cu will appear. The volume ratio of one mole Cu2O to two mole copper is calculated in Eq. (2).

V ð1 mol Cu2 OÞ ¼ 1:459 V ð2 mol CuÞ

ð2Þ

In as much as the volume ratio is more than one, pullout of Cu2O particles results in an increase of the pore density on the interface. Copper side fracture surfaces corresponding to different stages are shown in Fig. 7. In contrast to first stage, the pores sizes are diminished in the second stage. On account of the smaller pores Fig. 5. The peel strength of the second stage specimens.

Fig. 6. The maximum peel strength in the first and second stages.

Moreover, the maximum peel strength of specimens increased to 17.1 ± 0.2 kg/cm corresponding to a new optimum value of the CuO layer thickness. Increasing peel strength of the eutectically bonded interface is attributed to the removal of Cu2O particles from the interface. After the solidification of interfacial liquid, there is a distribution of Cu2O particles in interface. If reducing atmosphere exists, the Cu2O particles will reduce to Cu. The removal of Cu2O particles from the interface results in the substitution of Cu2O/Al2O3 and Cu2O/Cu interfaces by Cu/Al2O3 interface. According to Chiang et al. and Sun and Driscoll investigations [4,12], the strength of Cu/Al2O3 is more than Cu2O/Cu; hence, improved alumina–copper contact surface and decreasing stress concentration owing to absence of Cu2O particles resulted in the increase of the peel strength. Three main factors determine the peel strength of the joint; wetting properties of the copper–oxygen liquid on the alumina surface, Cu2O particles density and distribution, and pores. Yoshino reported that pores detrimental effect on peel strength is low comparing the wetting effect [2]. Since in the second stage Cu2O particles have been removed, wetting parameter remains the dominant parameter on peel strength. It has been proved [9] that reaching a threshold value for the oxygen content within the copper–oxygen liquid phase, the wetting angle do not change considerably. The observed decrease of the peel strength for high thickness of CuO layers after reaching the optimum peel strength, as shown in

Fig. 7. The copper side fracture surfaces of three stages.

1102

H. Ghasemi et al. / Materials and Design 30 (2009) 1098–1102

in vacuum furnace environment, the pressure difference between the furnace environment and released O2 (gas) is high; O2 molecules can extend to larger dimensions. Hence, it can be said that the furnace environment gas does not have significant effect on pores formation. In a conclusion, the pores on alumina–copper eutectic bond can be formed due to two reasons. O2 gas release is the first one. Pullout of Cu2O particles is the second one. However, in some pores both conditions were prepared. In these pores, a portion of pore is formed due to gas release and in perimeter of these pores Cu2O particles are present. These Cu2O particles are responsible for irregular shape of these kinds of pores. 4. Conclusion

Fig. 8. The comparison between maximum peel strength in three stages.

in interface, the tolerances in peel strength has diminished which is clear in Fig. 6. By precise focus, it can be found that smaller spherical pores are same in first and second stages. However, larger pores in the first stage which have irregular shape are contracted to smaller irregular pores in the second stage. According to Sun and Driscoll investigations [12], the Cu2O content of interface layer after solidification is more than Cu2O content of eutectic point. (The Cu2O content of eutectic point is 4.3 vol%.). Moreover, this fact is clear in pictures which other researchers have taken from the alumina–copper bond [5,6,8]. So there are some Cu2O particles in liquid copper in bonding temperature. These Cu2O particles can be reduced at bonding temperature in 2  106 atm oxygen partial pressure. Reduction of Cu2O particles forms O2 (gas) molecules. These molecules can form spherical pores at interface. But this reduction reaction has not completed during the bonding process. The Cu2O particles in interface after solidification are evidence to this conclusion. Thus, until now, there are two sources for pores formation. First one is the release of O2 molecules formed due to reduction of Cu2O particles in liquid copper. Pullout of Cu2O particles from copper fracture side is the second one. Comparing the Cu2O particles on alumina fracture surface with irregular pores size, it is clear that Cu2O particles are smaller than irregular pores on copper fracture side. So, maybe the irregular pores on fracture surface of copper are formed due to several sources. In view of the fact that furnace atmosphere gas was predicted as one of sources for pores formation [6], In the hope of removing some pores, the final stage of bonding was carried out at vacuum furnace. The vacuum atmosphere was 5  108 atm. The maximum peel strength of this stage is illustrated in Fig. 8. As depicted in Fig. 8, the peel strength has decreased and its tolerances are more in comparison with the second stage. In this stage, same as second stage, the interface is free of Cu2O particles as it was predicted due to suitable environment for Cu2O particles reduction. The outstanding point in the third stage is the formation of larger pores at the interface (Fig. 7). Furthermore, the shape of pores has changed. Drop in peel strength is related to formation of larger pores, and these pores are cause of more tolerances in peel strength. Formation of larger pores can be attributed to furnace atmosphere. There are two reactions which are responsible for larger pores. The reduction of Cu2O particles in liquid copper is the first one; presence of vacuum condition is the second one. As it was mentioned before, there are some Cu2O particles in liquid copper in bonding temperature. Since the furnace oxygen partial pressure is approximately 108 atm, these oxide particles reduce more than two previous stages and liberate more O2 gas. Also, Duo to bonding

Examinations on strength of eutectically bonded Cu–Al2O3 show that under 2  106 atm oxygen partial pressure, optimum CuO layer thickness, by preoxidation method, is 320 ± 25 nm which results in 13.1 ± 0.3 kg/cm peel strength. To achieve, more strength, Cu2O particles can be removed from eutectic interface by injecting H2 gas into the furnace. In this case, the peel strength is raised up to 17.1 ± 0.2 kg/cm; microstructure investigations illustrate direct contact of Cu to Al2O3 without any interfacial phase. On sources of pores formation at the interface, it is shown that furnace noble gas does not have a considerable effect on strength. Besides, performing the bonding procedure in a vacuum furnace decreases the strength due to formation of larger pores and causes more fluctuations in bond strength. Finally, it has been demonstrated that pores at the interface of eutectically bonded Cu–Al2O3 are due to two main reasons: O2 release in bonding temperature and pullout of Cu2O particles. Acknowledgements Financial support for this work was provided by Bonab Research Center. The assistance of the following people in various aspects of the experimental work is greatly appreciated: Mr.T.Tohidi, Mr. M. Naderi. References [1] Burgess JF, Neugebauer CA. Direct bonding of metals with a metal-gas eutectic. US Pat. No. 3854892; 1974. [2] Yuichi Yoshino. Role of oxygen in bonding copper to alumina. J Am Ceram Soc 1989;72(8):1322–7. [3] Yoshino Y, Ohtsu H. Interface structure and bond strength of copper-bonded alumina substrates. J Am Ceram Soc 1991;74:2184–8. [4] Chiang WL, Greenhut VA, Shanefield DJ, Johnson LA, Moore RL. Gas–metal eutectic bonded Cu to Al2O3 substrate-mechanism and substrate additives effect study. Ceram Eng Sci Proc 1993;14(9–10):802–12. [5] Sung Tae Kim, Chong Hee Kim. Interfacial reaction product and its effect on the strength of copper to alumina eutectic bonding. J Mater Sci 1992;27:2061–6. [6] Seager CW, Kokini K, Trumble K, Krane MJM. The influence of CuAlO2 on the strength of eutectically bonded Cu/Al2O3 interfaces. Scripta Mater 2002;46:395–400. [7] Trumble KP. Thermodynamic analysis of aluminate formation at Fe/Al2O3 and Cu/Al2O3. Acta Metall Mater 1992;40:S105–10. [8] Ning H, Ma J, Huang F, Wang Y. Preoxidation of the Cu layer in direct bonding technology. Appl Surf Sci 2003;211:250–8. [9] Matthias Diemer, Achim Neubrand, Kevin P Trumble, Jurgen Rodel. Influence of oxygen partial pressure and oxygen content on the wettability in the copper– oxygen–alumina system. J Am Ceram Soc 1999;82(10):2825–32. [10] Reimanis IE, Trumble KP, Rogers KA, Dalgleish BJ. Influence of the Cu2O and CuAlO2 interphases on crack propagation at Cu/a–Al2O3 interfaces. J Am Ceram Soc 1997;80(2):424–32. [11] Haider J, Rahman M, Corcoran B, Hashmi MSJ. Simulation of thermal stress in magnetron sputtered thin coating by finite element analysis. J Mater Process Technol 2005;168:36–41. [12] Sun YS, Driscoll JC. A new hybrid power technique utilizing a direct copper to ceramic bond. IEEE Trans Electr Dev 1976;ED-23(8):961–7.

Roles-of-preoxidation-Cu2O-particles-and-interface-pores-on-the ...

... of alumina by liquid copper. The noteworthy point in Fig. 3a. and 3b is the rising and falling rate of peel strength versus the. oxide layer thickness. It shows that ...

614KB Sizes 0 Downloads 286 Views

Recommend Documents

No documents