Transaction B: Mechanical EngineeringVol. 16, No. 3, pp. 263{268c Sharif University of Technology, June 2009

Alumina-Copper Eutectic BondStrength: Contribution of Preoxidation,

Cuprous Oxides Particles and Pores

H. Ghasemi1;�, M.A. Faghihi Sani1, A.H. Kokabi1 and Z. Riazi2

Abstract. The in uences of cupric oxide layer thickness, cuprous oxide particles and pores on themechanical properties and microstructure of an alumina-copper eutectic bond have been investigated. Thefurnace atmosphere in the �rst stage was argon gas with 2 � 10�6 atm oxygen partial pressure. In thesecond stage, the furnace atmosphere was the same as the �rst stage except that the cooling interval wasbetween 900-1000�C and the hydrogen gas was injected into the furnace atmosphere. Finally, in the laststage, a vacuum furnace with 5 � 10�8 atm pressure was chosen for the bonding procedure. The peelstrength of �rst stage specimens shows that a cupric oxide layer with 320 � 25 nm thickness generatesmaximum peel strength (13:1� 0:3 kg/cm) in the joint interface. In the second stage, by using hydrogengas, a joint interface free from any cuprous oxide particles was formed. In this case, the joint strengthhas increased to 17:1 � 0:2 kg/cm. Finally, the bonding process in the vacuum furnace indicates thatthe furnace gas does not have a considerable e�ect on joint interface pores. Furthermore, the bondingprocess in the vacuum furnace reduces the peel strength of the joint due to the formation of more pores.A thorough study of pore formation is presented.

Keywords: Alumina-copper; Bonding; Peel strength.

INTRODUCTION

One of the methods for joining copper and alumina isgas-metal eutectic bonding. The bonding of copper toalumina is carried out at 1075�2�C in a properly con-trolled oxygen-containing atmosphere [1]. The role ofoxygen is to improve the wetting process of alumina byliquid copper in order to achieve a homogeneous, liquidcopper �lm spreading along the bonding interface [2].So far, the mechanical properties of eutectic bondingand corresponding interactions within the bonding areahave been studied extensively by several authors [2-8]. Yoshino [2] observed that the peel strength ofan alumina-copper bond is predominantly a�ected bydissolved-oxygen concentration. He reported that thecuprous oxides and voids also contribute to the truebond strength. Oxygen seems to a�ect the peel

1. Department of Materials Science and Engineering, SharifUniversity of Technology, Tehran, Iran.

2. Bonab Research Center, Bonab.*. Corresponding author. E-mail: hadighasemi@alum.sharif.

edu

Received 25 June 2007; received in revised form 26 December2007; accepted 26 May 2008

strength of the alumina/copper bond, principally intwo ways: Through a change in cohesive bond energyand through the formation of cuprous oxide. The mea-sured bond strength is the combination of these e�ects.Yoshino predicted that by removing the oxide particlesfrom the interface the peel strength of the joint can beincreased to 15 kg/cm. Moreover, Trumble et al. [7]have reported that Cu2O particle concentration canbe reduced after bonding without any loss in bondstrength. Ning [8] showed that a threshold thicknessof the cupric oxide layer generates the highest bondingstrength. This threshold thickness is a function ofthe alumina substrate pro�le. On the source of poreformation and pore e�ects on the peel strength ofthe eutectic bond, Yoshino [2] reported that the poreswithin the bonding interface can be formed due to therelease of oxygen gas in the liquid copper. Therefore,increasing the thickness of the cupric oxide layer willresult in larger pore size. However, Seager et al. [6]found that smaller pores (1-3 �m) may be pullouts ofthe Cu2O particles observed on the alumina fracturesurface, while the large pores observed by Yoshinoare due to argon entrapment in liquid copper duringprocessing. Concerning the in uence of pores on the

264 H. Ghasemi, M.A. Faghihi Sani, A.H. Kokabi and Z. Riazi

mechanical strength of eutectic bond, Reimanis [9] haveobserved that crack-front perturbation occurs when thecrack tip is in contact with a pore; the crack frontis drawn into the pore which causes the debonding ofthe regions immediately surrounding the pore. Despiteseveral investigations, the accurate contribution ofpreoxidation, cuprous oxide particles and pores on thepeel strength of this bond is unclear. The goal of thecurrent research is to shed more light on the e�ects ofcupric oxide layer thickness, cuprous oxides and poredensity, size and distribution on the peel strength of analumina-copper eutectic bond. Besides, the sources ofpore formation have been investigated thoroughly.

EXPERIMENTAL SETUP ANDMETHODOLOGY

Experiments were carried out in three stages. In the�rst stage, the e�ect of preoxidation on peel strengthof alumina-copper bond has been examined. In thesecond stage, the in uence of cuprous oxide particles onthe peel strength of an alumina-copper bond has beeninvestigated. Finally, in order to decrease the interfacepores, the bonding process was performed in a vacuumfurnace. All of the alumina-copper bonds began with99.99% Cu strips and 97% Al2O3. Copper strips andalumina 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, thestrips were degreased by scrubbing the joint surfacesin a solution of liquid detergent, washed with cleanhot water and dried thoroughly in a steam of hot air.Secondly, the strips were immersed in 25% nitric acidfor 30s. Thirdly, the strips were dipped in ethyl alcoholwashed by clean cold water and dried by hot air. Onthe other hand, alumina specimens were polished bydiamond paste with an average particle size of 0.5 �m,cleaned by ultrasonic treatment in ethyl alcohol for15 min and rinsed in distilled water. At the �nal step,alumina specimens were annealed in air at 1000�C toeliminate any hydroxyl groups.

Copper strips were preoxidized at various temper-atures and several times to obtain a range of cupricoxide (CuO) thickness. On account of the phasestability diagram of copper oxidation, the formed oxidephase below 400�C under atmospheric pressure willbe CuO [10]. The thickness of the cupric oxide layerwas calculated from the weight di�erence before andafter preoxidation. In the �rst stage, the bondingprocess was carried out in a tube furnace under anargon atmosphere (< 2 ppm O2). The gas ows forheating and cooling procedures were respectively, 150and 200 L/hr. The samples were heated at a rate of10�C/min to 1000�C, followed by a rate of 2�C/minto 1075�C, where they were held for 1 hr. Duringthe holding time, an alumina-copper bond was formed.

After 1 hr, the temperature was decreased slowly(5�C/min) to 700�C and then 10�C/min to 400�C.The temperature was held for 1 hr at 400�C to reducethermal stresses. Finally, the furnace cooled to roomtemperature. The furnace temperature curve is shownin Figure 1. In the second stage, the experimentalprocedure for bond formation was identical to the �rststage, except for the cooling interval which was between900-1000�C in which hydrogen gas was injected intothe furnace atmosphere. In the last stage, the bondingprocess was carried out in a vacuum furnace. A vacuumtube with a mechanical pump and an oil di�usion pumpcan achieve a vacuum atmosphere of 5 � 10�8 atm.The oxygen partial pressure was measured by a CaO-stabilized ZrO2 oxygen sensor. The peel strength of thebonded specimens was measured by an Instron MachineModel 1115. Fracture surfaces were analyzed by opticalmicroscopy and scanning electron microscopy.

RESULTS AND DISCUSSION

Cupric Oxide Layer Thikness

The thickness of the cupric oxide layer, formed underdi�erent conditions of temperature and oxidation time,is shown in Figure 2. It shows that the thicknessvariation of the cupric oxide layer is gradual anduniform within the range of 0 to 600 nm above whichthe thickness increase is abrupt and discontinuous.Samples having an oxide layer thickness greater than600 nm present a discontinuous and cracked oxide layersurface. Furthermore, the cohesion between the cupricoxide layer and the copper strips has decreased notably.These discontinuities on the cupric oxide layer surfaceare due to the thickness of the CuO layer. The morethe oxide layer thickness, the more the sensitivity toa thermal expansion coe�cient mismatch. Since thethermal expansion coe�cients of Cu (20 � 10�6�C�1)and CuO (4:3�10�6 �C�1) have a great di�erence, thestresses formed in cooling are high. The thermal stress

Figure 1. The furnace curve in bonding process.

Alumina-Copper Eutectic Bond Strength 265

�eld generates discontinuities and cracks within thecupric oxide layer, which in turn results in an increaseof the copper oxidation rate, due to a greater copperexposure. Hence, the preoxidation conditions, whichdo create more than, approximately, a 500-600 nmoxide layer, result in unsuitable strips for the bondingprocess.

Peel Strength Measurments

The peel strengths of �rst stage specimens are shownin Figure 3. As shown in Figure 3a, an optimumoxide layer thickness results in the highest peel strength(13:1�0:3 kg/cm) under a 2�10�6 atm oxygen partialpressure. This optimum thickness is a result of theinteraction between the wetting of the alumina surfaceand the percentage of oxide particles at the interface.Thicker Cupric oxide layers introduce more oxygen intothe interface and, subsequently, the wetting angle willbe decreased [11] and more cuprous oxide particles(Cu2O) will remain in the interface after bonding.Similarly, Figure 3b shows an optimum point. By com-paring the optimum layer thickness in Figures 3a and3b, a di�erence of approximately 25 nm is observed.This di�erence can be resulted from errors in thepeel strength and oxide layer thickness measurement.Thus, it can be said that 320� 25 nm is the optimumoxide layer thickness for alumina-copper bonding under2�10�6 atm oxygen partial pressure. The peel strengthdi�erence in these two states is low; it shows thatoxidation at di�erent temperatures, to obtain a certainoxide layer thickness, does not a�ect the peel strength.Both Figures 3a and 3b reveal that, before reaching theoptimum point, the peel strength of both specimenshas raised steadily. It can be concluded that improvedwetting, due to an increase in the oxygen content ofthe liquid copper, has heightened the peel strength.After an optimum point, the e�ect of the Cu2O contentpredominates, decreasing the bond strength. Figure 3c

Figure 2. The variation of oxide layer thickness indi�erent preoxidation conditions.

shows the peel strength of specimens oxidized at 400�C.In contrast to Figures 3a and 3b, in the preoxidationtime span of this temperature, the peel strength ofspecimens has decreased slowly. It means that oneof the factors a�ecting peel strength is predominant.Obviously, in this time span, the Cu2O content ofthe interface layer is the predominant parameter tothe wetting of the alumina by liquid copper. Theimportant point in Figures 3a and 3b is the rising andfalling rate of the peel strength versus the oxide layerthickness. It shows that the e�ect of wetting on peelstrength is more signi�cant than the e�ect of the Cu2Ocontent of the interface. This state is clear in the Ninginvestigations [8].

To achieve more strength in the alumina-copperbond, it is desirable to remove cuprous oxides fromthe interface. This fact is due to the higher strengthof the Cu/Al2O3 bond compared with Cu/Cu2O andCu2O/Al2O3 [2,4,7]. Hence, during the cooling intervalof 900-1000�C within the second stage, the hydrogengas was injected into the furnace atmosphere in orderto remove Cu2O particles from the eutectic zone ofthe interface. Figure 4 shows the cross section of thealumina-copper joint at the second stage. It surpris-ingly shows no trace of the interface layer; the joint isdirectly established between alumina and copper. Inaddition to copper and alumina, no sign of any otherphase is detectable at the fracture surfaces. The peelstrength measurements of the joint at this stage areshown in Figure 5.

In addition, Figure 6 depicts the force versusdistance curve of the maximum peel strength at the�rst and second stages. As shown in Figure 5, incontrast to �rst stage specimens, the peel strengthof the joint has raised considerably, regardless of theoxide layer thicknesses. Moreover, the maximum peelstrength of specimens increased to 17:1 � 0:2 kg/cmcorresponding to a new optimum value of the cupricoxide layer thickness. The increasing peel strength ofthe eutectically bonded interface is attributed to theremoval of Cu2O particles from the interface. Afterthe solidi�cation of the interfacial liquid, there is adistribution of Cu2O particles in the interface. If areducing atmosphere exists, the Cu2O particles willreduce to Cu. The removal of Cu2O particles fromthe interface results in the substitution of Cu2O/Al2O3and Cu2O/Cu interfaces by a Cu/Al2O3 interface.According to Chiang et al. and Sun and Driscollinvestigations [4,12], the strength of the Cu/Al2O3interface is greater than that of the Cu2O/Cu interface.Therefore, an improved alumina-copper contact surfaceand a decreasing stress concentration, owing to theabsence of Cu2O particles, resulted in the increase ofthe peel strength.

Three main factors determine the peel strengthof the joint: Wetting properties of the copper-oxygen

266 H. Ghasemi, M.A. Faghihi Sani, A.H. Kokabi and Z. Riazi

Figure 3. The peel strength of specimens preoxidized indi�erent conditions.

liquid on the alumina surface, Cu2O particle densityand distribution, and pores. Yoshino reported thatthe pores detrimental e�ect on the peel strength is lowcompared to the wetting e�ect [2]. Since, at the secondstage, Cu2O particles have been removed, the wettingparameter remains the dominant parameter on peelstrength. It has been proved [11,13] that by reachinga threshold value for the oxygen content within thecopper-oxygen liquid phase, the wetting angle does notchange considerably. The observed decrease of peel

Figure 4. The cross section of alumina-copper joint afterremoving the Cu2O particles.

Figure 5. The peel strength of the second stagespecimens.

Figure 6. The maximum peel strength in the �rst andsecond stages.

strength for high thicknesses of cupric oxide layers afterreaching optimum peel strength, as shown in Figure 5,is due to an increase in pore density on the bondinginterface.

The change of the CuO optimum layer thicknessat this stage was predicted. Due to a reduction inCu2O particles, the e�ect of wetting is more signi�cant

Alumina-Copper Eutectic Bond Strength 267

than that of an increase in pore concentration anddistribution for the increasing oxygen content of theliquid copper, resulting in a shift of the optimum peelstrength to the values corresponding to a thicker CuOlayer. By reducing one mole of Cu2O, two mole of Cuwill appear. The volume ratio of one mole of Cu2O totwo mole copper is calculated in the following equation:

V (1 mol Cu2O)V (2 mol Cu)

= 1:459: (1)

Since the volume ratio is more than one, the pulloutof Cu2O particles results in an increase of the poredensity on the interface. Copper side fracture sur-faces, corresponding to di�erent stages, are shown inFigure 7. In contrast to the �rst stage, the poressizes are diminished in the second stage. On accountof the smaller pores in the interface, the tolerancein peel strength has diminished, which is clear inFigure 6. Paying close attention, it can be found thatsmaller spherical pores are the same at the two stages.However, larger pores at the �rst stage, which havean irregular shape, are contracted to smaller irregularpores at the second stage.

According to Sun & Driscoll investigations [12],the Cu2O content of the interface layer, after so-lidi�cation, is more than the Cu2O content of theeutectic point (the Cu2O content of the eutectic point is4.3% vol.). Moreover, this fact is clear in pictures thatother researchers have taken from the alumina-copperbond [5,6,8]. So, there are some Cu2O particles inliquid copper in the bonding temperature. These Cu2Oparticles can be reduced at the bonding temperaturein a 2 � 10�6 atm oxygen partial pressure. Thereduction of Cu2O particles forms O2 (gas) molecules.These molecules can form spherical pores at the in-terface. But, this reduction reaction is not completedduring the bonding process. The Cu2O particles inthe interface after solidi�cation are evidence of thisconclusion. Thus, until now there are two sources ofpore formation. The �rst is the O2 molecules thatare formed due to the reduction of Cu2O particles inthe liquid copper. The pullout of Cu2O particles fromthe copper fracture side is the second. Comparing theCu2O particles on the alumina fracture surface withirregular pore sizes, it is clear that Cu2O particles aresmaller than irregular pores on the copper fracture side.So, maybe the irregular pores on the fracture surfaceof copper are formed due to several sources.

In view of the fact that furnace atmosphere gaswas predicted as being one of the sources of poreformation [6], in the hope of removing some pores,the �nal stage of bonding was carried out in a vacuumfurnace. The vacuum atmosphere was 5 � 10�8 atm.The maximum peel strength of this stage is illustratedin Figure 8. As depicted in Figure 8, the peelstrength has decreased and its tolerances are more,

Figure 7. The copper side fracture surfaces of threestages.

in comparison with the second stage. At this stage,as at the second stage, the interface is free of Cu2Oparticles, as predicted, due to a suitable environmentfor Cu2O particle reduction. The outstanding point atthe third stage is the formation of larger pores on theinterface (Figure 7). Furthermore, the shape of themhas changed. The drop in peel strength is related tothe formation of larger pores, and these pores are thecause of more tolerances in peel strength.

The formation of larger pores can be attributedto the furnace atmosphere. There are two reactionsresponsible for larger pores: The reduction of Cu2Oparticles in liquid copper and the presence of a vacuumcondition. As mentioned before, there are some Cu2Oparticles in liquid copper in the bonding temperature.Since the furnace oxygen partial pressure is 10�8 atm,these oxide particles reduce more than at the two

268 H. Ghasemi, M.A. Faghihi Sani, A.H. Kokabi and Z. Riazi

Figure 8. The comparison between maximum peelstrength in three stages.

previous stages and liberate more O2 gas. Also, dueto bonding in the vacuum furnace environment, thepressure di�erence between the furnace environmentand the released O2 (gas) is high; O2 molecules canextend to larger dimensions. Therefore, it can besaid that the furnace environment gas does not havea signi�cant e�ect on pore formation.

In conclusion, the pores on the alumina-coppereutectic bond can be formed due to two reasons: O2 gasrelease, and the pullout of Cu2O particles. However,in some pores both conditions were prepared. In thesepores, a portion of the pore is formed due to gas releaseand, at the perimeter of these pores, Cu2O particles arepresent. These Cu2O particles are responsible for theirregular shape of these kinds of pore.

CONCLUSION

1. Creating more than 500-600 nm cuprous oxidelayer thickness by preoxidation makes the stripsimproper for eutectic bonding due to the formingof cracks in the oxide layer and a loss of adhesion.

2. The optimum cuprous oxide layer thickness forproviding the highest peel strength in an alumina-copper eutectic joint, in 2 � 10�6 oxygen partialpressure, is 320 � 25 nm, which results to 13:1 �0:3 kg/cm peel strength.

3. By removing all of the Cu2O particles from thealumina-copper interface, the peel strength hasrisen to 17:1� 0:2 kg/cm.

4. Furnace gases do not have a signi�cant e�ect onpore formation in an alumina-copper interface.

5. Performing the alumina-copper eutectic bond in avacuum furnace decreases the peel strength andcauses more uctuations in joint strength due tothe formation of larger pores.

6. The pores at the alumina-copper interface areformed due to two reasons: O2 gas release, and thepullout of Cu2O particles.

ACKNOWLEDGMENTS

Financial support for this work was provided by theBonab Research Center. The assistance of the followingorganizations and people regarding various aspects ofthe experimental work is greatly appreciated: theResearch Vice-Presidency of SUT, Mr. T. Tohidi andMr. M. Naderi.

REFERENCES

1. Burgess, J.F. and Neugebauer, C.A. \Direct bondingof metals with a metal-gas eutectic", U.S.Pat., 3854892(1974).

2. Yoshino, Y. \Role of oxygen in bonding copper toalumina", J. Am. Ceram. Soc., 72(8), pp. 1322-1327(1989).

3. Yoshino, Y. and Ohtsu, H. \Interface structure andbond strength of copper-bonded alumina substrates",J. Am. Ceram. Soc., 74, pp. 2184-2188 (1991).

4. Chiang, W.L., Greenhut, V.A., Shane�eld, D.J., John-son, L.A. and Moore, R.L. \Gas-metal eutectic bondedCu to Al2O3 substrate-mechanism and substrate addi-tives e�ect study", Ceram. Eng. Sci. Proc., 14(9-10),pp. 802-812 (1993).

5. Kim, S.T. and Kim, C.H. \Interfacial reaction productand its e�ect on the strength of copper to aluminaeutectic bonding", J. Materials Science, 27, pp. 2061-2066 (1992).

6. Seager, C.W., Kokini, K., Trumble, K. and Krane,M.J.M. \The in uence of CuAlO2 on the strengthof eutectically bonded Cu/Al2O3 interfaces", ScriptaMaterialia, 46, pp. 395-400 (2002).

7. Trumble, K.P. \Thermodynamic analysis of aluminateformation at Fe/Al2O3 and Cu/Al2O3", Acta Metall.Mater., 40, pp. S105-S110 (1992).

8. Ning, H., Ma, J., Huang, F. and Wang, Y. \Preoxi-dation of the Cu layer in direct bonding technology",Applied Surface Science, 211, pp. 250-258 (2003).

9. Reimanis, I.E., Trumble, K.P., Rogers, K.A. andDalgleish, B.J. \In uence of the Cu2O and CuAlO2

interphases on crack propagation at Cu/�-Al2O3 inter-faces", J. Am. Ceram. Soc., 80(2), pp. 424-432 (1997).

10. Kubaschewski, O and Hopkins, B.E. \Oxidation ofmetals and alloys", Butterworths Publications (1967).

11. Diemer, M., Neubrand, A., Trumble, K.P. and Rodel,J. \In uence of oxygen partial pressure and oxy-gen content on the wettability in the copper-oxygen-alumina system", J. Am. Ceram. Soc., 82(10), pp.2825-2832 (1999).

12. Sun, Y.S. and Driscoll, J.C. \A new hybrid powertechnique utilizing a direct copper to ceramic bond",IEEE Trans. Electron Devices, ED-23(8), pp. 961-967(1976).

13. Lewinsohn, Ch.A., Singh, M. and Loehman, R. \Ad-vances in joining of ceramics", J. American CeramicSociety, 90, pp. 77 (2003).