Iranian Journal of Materials Science and Engineering, Vol. 4, Numbers 3 and 4, Summer and Autumn 2007

AN INFLUENCE OF PHASE DEVELOPMENT ON MECHANICAL STRENGTH OF ALUMINA-COPPER JOINT PREPARED BY MOLYMn METHOD H. Ghasemi 1, M. A. Faghihi Sani 1 and Z. Riazi 2 [email protected] 1 2

Department of materials science and engineering, Sharif University of Technology, Tehran, Iran Bonab Research Center, Bonab, Iran Abstract: The effect of phase development on peel strength of alumina-copper metalized joint has been investigated. The alumina-copper joint was prepared in three stages. The alumina substrate was, first, metalized at 1500°C in H2-furnace by a new formulation. In the second step, a nickel layer was electroplated on the metalized layer with approximately 10µm thickness. Finally, copper strips were bonded to metalized alumina with Ag-Cu (72-28) filler metal. The peel strength of the joint was 9.5±0.5 Kg/cm which shows approximately 30% increase in comparison to previous works. By study of fracture surface and crack propagation path, it has been concluded that this increase is due to the formation of more spinel phase. Keywords: Alumina-copper joint, Metalization, Spinel phase, Peel strength

1. INTRODUCTION Joining of metals to ceramics has been widely practiced in fabricating structural components to utilize the favorable characteristics of the engineering ceramics, such as high strength, high hardness, high thermal resistance, and so forth [1]. The ceramic-metal joining technologies used today range from simple mechanical attachments such as the compression fit used in spark plugs through to liquid phase processes such as adhesive bonding and brazing [2]. Among ceramic-metal joints, the alumina-copper joint is of great significance. Electronic boards, Vacuum tubes and catalysts are some instances of its application. Until now, Alumina- copper joint has been established by more than ten methods [3]. Among them, Moly-Mn process for its reliability is predominant process and is commonly used process in electronic industry. Moly-Mn process involves three distinct stages: metalizing, coating of a nickel layer, and brazing. In the first stage, a paste consisting of tungsten or molybdenum particles (due to their low thermal expansion coefficient), manganese oxide and glass phase formers is applied to the ceramic and sintered under the reducing conditions at 1500 °C [4,5]. Manganese oxide decreases the melting point and reduces the

14

viscosity of the glassy phase while increasing its total volume to enhance bonding [6]. The glassy phase of the ceramic infiltrates the metal particles near the surface creating the bond. When the ceramic will not provide sufficient glass, either glass frit or glass former is added to paste formulation. At the top surface, the metal particles are reduced and sintered to direct contact. The surface structure is virtually continuous metal. In the Second stage, to improve wetting of the metalized layer by brazing filler metal, a nickel layer is coated on metalized surface. Finally, the joint will be established between copper and metalized alumina by Ag-Cu filler metal. The final structure of the Moly-Mn joint provides a gradient in properties such as thermal expansion and elastic modulus, reducing local stress intensity [5]. Up to now, some formulas for making the metalized layer has been presented [4]. In some studies the presence of spinel phase either MnAl2O4 or MgAl2O4 has not been reported and in others the slight percentage of spinel phase has been observed [4, 6, 7, 8]. The peel strength of joints prepared by these formulas varies between 6-8 kg/cm [4, 5, 9, 10]. Even though numerous researchers have worked on this process, the effect of phase development particularly spinel phase on peel strength of joint

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

is unknown. Therefore, in current research by introducing more spinel phase in interface, the influence of phase development on the mechanical strength of the alumina-copper joint has been investigated. 2. EXPERIMENTAL PROCEDURE All of the copper-alumina bonding processes began with strips of 99.99% Cu and 97% Al2O3 which contains few percentages of glass phases. Copper strips (OF-Cu) and alumina specimens were 350 mm and 200 mm in length, 25 mm in width, and 0.8 mm and 1.5 mm in thickness, respectively. Surface cleaning of copper strips was carried out according to ISO 4588 standard. In other words, the joint surfaces were scrubbed in a solution of liquid non-ionic detergent, washed with clean hot water, and allowed to dry thoroughly in a steam of hot air; afterwards, the joint surfaces were wet abraded with 800 grit SiC abrasive paper and washed with ethyl alcohol and dried by hot air. Alumina specimens were, on the other hand, polished by diamond paste with average 0.5 µm particle size. The cleaning process was continued by ultrasonic washing and ethyl alcohol degreasing. The alumina specimens, in last step, were soaked in boiling distilled water for 10 min, and heat treated in an oxidizing furnace at 1000 °C for 1 hour to put its surface in proper condition for joining process [3]. Bonding process by MolyMn method was performed in three stages. First of all, a powder by mentioned formula in Table 1 was prepared. More MgO and Al2O3 particles was added to powder formula in contrast to previous formulas [4] to increase the spinel (MgAl2O4) content of joint interface. After mixing the powders in jarmill, they were blended with 10% nitrocotton to form a paste. The paste was applied on alumina surface homogeneously by a special tool. The paste was sintered on alumina substrate in H2-furnace; in this step, the specimens were heated at a rate of 20 °C/min up to 1500 °C, where they were held for 1 hr. After 1 hr, the temperature was decreased slowly to reduce the residual stresses (20 °C/min to 700 °C and then 50 °C/min to 400 °C). The temperature was then held for 2 hr at 400 °C, after which the furnace cooled to room temperature. In second step, a nickel layer with 10 µm thicknesses was coated on metalized

surface by electroplating process to make a surface with proper wetting behavior by filler metal and to prevent the detrimental reaction between metalized layer and filler metal. In last stage, the copper strips were bonded to alumina specimens with Ag-Cu (72-28) filler metal by heating in H2-furnace at 850 °C for 1hr [5]. After bonding process, five peel test specimens, according to ISO 8510-2 standard (Fig. 1), were prepared and their peel strengths were measured by Instron Universal Machine Model 1115 [11]. Finally, the fracture surfaces were analyzed by scanning electron microscope (SEM), EDX map, and X-ray diffraction (XRD) to distinguish the formed phases in microstructure and their effects on mechanical strength of alumina-copper joint. Table 1. Metalized powder composition. Component wt%

Mo

SiO2 MnO Al2O3 MgO

Other oxides

~ 50% ~ 7 % ~ 8 % ~ 20 % ~ 10 % 5 %

Particle size 0.5-1 µm

1-2 µm

Fig. 1. The peel test sample (ISO 8510-2).

15

Iranian Journal of Materials Science and Engineering, Vol. 4, Numbers 3 and 4, Summer and Autumn 2007

3. RESULTS AND DISCUSSION 3.1. Strength Measurment Fig. 2 depicts the force versus distance curve in peel strength test. The peel strength is introduced as force per unit width necessary to rupture a joint. Thus, the peel strength can be calculated from the division of the force on specimen width (2.5 cm). The average peel strength of the joint is 9.5±0.5 Kg/cm. The variation along the bonding interface is negligible as shown in Fig. 2. So, it can be concluded that a homogenous bond has been formed at interface. Also, the force variation in Fig. 2 illustrate that in some areas, bond strength has risen sharply and after that has fallen. It means that in these areas there are some obstacles on crack propagation. Comparing the mechanical results with previous works [4,5,9,10] indicates the 30% increase in joint peel strength. This increase is likely due to formed phases and their distribution in the interface.

Fig. 3. Optical microscopy of cross section of the joint.

0.3

0.25

Force (KN)

0.2

0.15

Fig. 4. Distribution of phases in metalized layer. Metalized Layer X-Ray Diffraction

0.1

Mo = Molybdenum , MA = MgAl2O4 , A= Al2O3 1400

Mo

0.05 1200

0 2

4

6

8

10

12

14

1000

16

Distance (Cm)

Fig. 2. The force versus distance curve in peel strength.

Intensity

0

Mo

800 600

Mo MA

Mo

400

3.2. Microstructural Evaluation The correlation between peel strength increase and microstructure has been investigated. Fig. 3 shows the cross section of the joint. As illustrated in Fig. 3, the bonding interface consists of three distinct layers, the metalized layer, the nickel layer, and the braze layer. The slight reactions have occurred between layers. Metalized layer, nickel layer, and braze layer are 25µm, 10µm, and 40µm in thickness, respectively. The metalized layer is magnified in Fig. 4. The phases in this layer characterized by X-ray Diffraction (XRD) (Fig. 5) and EDX map are marked in Fig. 4.

16

MA

MA 200

MA

A

MA A

MA A

MA

0 10

20

30

40

50

60

70

80

90

100

Diffraction Angle

Fig. 5. Phases in metalized layer.

Fig. 4 shows the Mo particles which have formed approximately 50% of metalized layer. These particles have been spread out throughout the metalized layer, and have formed a contiguous network. In contrast with previous works [4, 6, 7] which includes a distinct glass phase with dispersed Mo Particles, there is no evidence of a distinct glass layer. A spinel layer has formed on alumina substrate and serves as a

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

connection between alumina substrate and Mo continuous network. Due to inadequate glassy phase some pores was unfilled with glassy phase. To investigate the metalized layer in depth, EDX map of four main elements in this layer are illustrated in Fig.6. Si map in Fig. 6 shows the distribution of glass phase in metalized layer. As depicted in Fig. 7, the oxides contributing in glass phase formation are SiO2, MnO, Al2O3, and MgO. Si map distribution demonstrates that the glass phase has been dispersed throughout metalized layer inside pores and among other phases. There is no spot

Al map

Mo map

concentration of glass phase in metalized layer. According to XRD pattern (Fig. 5), Mo, MgAl2O4 and Al2O3 are crystalline phases in metalized layer. Hence, Simultaneous presence of Al and Mg elements (Al and Mg map in Fig. 6) implies existence of spinel phase. Some areas where Al concentration is not matched to Mg and Si concentration, Al2O3 must be formed. In other areas, Al and Mg elements have dissolved in glass phase. According to Belic work [8], glass phase formation in joint interface begins at 1191 °C.

Mg map

Si map

Fig. 6. EDX map of four main elements in the metalized layer.

17

Iranian Journal of Materials Science and Engineering, Vol. 4, Numbers 3 and 4, Summer and Autumn 2007

Fig. 8. Fracture surface (Copper side). Fig. 7. EDX of glass phase.

18

Fig. 9. Fracture surface (Alumina side). Fracture Surface X-Ray Diffraction (Copper Side) Mo = Molybdenum , C = Copper , N = Nickel , S = Silver 1400

Mo 1200

Intensity

1000 800

Mo 600 400

Mo

CNS 200

CNS

MA MA

Mo C NS

MA

N

0 10

20

30

40

50

60

70

80

90

100

Diffraction Angle

Fig. 10. Phases in fracture surface (Cop. side). Fracture Surface X-Ray Diffraction (Alumina Side) Mo = Molybdenum , MA = MgAl2O4 , A = Al2O3 1400

Mo 1200 1000

Intensity

Hence, at sintering temperature (1500 °C), there is a flowing liquid phase which easily flows between Mo, Al2O3, and MgO particles. In addition, due to capillarity effect, a portion of alumina’s grain boundary glass phase diffuse to joint interface and mix with mentioned liquid phase [4,8]. In 1500°C, a portion of alumina powder and alumina substrate, moreover, have reacted with MgO powder and have formed spinel phase. Although spinel formation cause stress in joint interface on account of its volume expansion, the existence of liquid phase in joint interface release stresses. At the same time, Mo particles are sintered together and have formed a continuous network through the joint. The interaction between Mo particles, glass phase, spinel, and alumina particles has led to bonding of metalized layer to alumina specimen. As mentioned before, Mo particles have created a continuous network. The glass phase formed in interface has flowed to some pores of this network and have formed a mechanical connection between Mo network and glass phase. According to Nascimento work [4], the main effect of spinel phase is connecting Mo particles to ceramic substrate since its expansion coefficient falls between the expansion coefficient of ceramic substrate and molybdenum. However, to discover the spinel phase other effects, more spinel phase is formed in current study inside the interface. The phases in each side of fracture surface (Fig. 8, Fig. 9) were identified by X-ray diffraction (XRD) method and are shown in Fig. 10 and Fig. 11.

800

A

600

Mo

A A

A

400

A

MA 200

A

MA

A

MA

Mo A MA MA

A MA

Mo

0 10

20

30

40

50

60

70

80

90

Diffraction Angle

Fig. 11. Phases in fracture surface (Alum. side).

100

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

Present phases in fracture surfaces, according to XRD patterns (Fig. 10 and Fig. 11), indicate that the fracture has occurred in metalized layer. There is no Al2O3 particles on fracture surface of copper (Fig. 10) and the percentage of MgAl2O4 in fracture surface is low. In contrast, the percentage of Al2O3 and MgAl2O4 particles on fracture surface of alumina side is high. This shows that majority of spinel and alumina particles in interface have formed close to the alumina substrate. The formed MgAl2O4 layer on alumina substrate in Fig. 4 is an evidence for this conclusion. Furthermore, this is a proof for Nascimento conclusion [4] which shows that

Al Map

Mo Map

spinel phase serve as an intermediate layer between molybdenum and ceramic substrate. In Fig. 8, it is obvious that brittle fracture has occurred in glassy areas. Yet, ductile fracture has occurred in Mo areas. Fracture surface of copper side have analyzed by EDX map (Fig. 12) to find the contribution of phases in joint rupture. According to Si map, the percentage of glass phase in fracture surface is high and glass phase is dispersed throughout the fracture surface. It means that crack has propagated in glass phase. In areas that Mg and Al map matched together without the presence of Si, spinel phase (MgAl2O4) has been formed.

Mg Map

Si Map

Fig. 12. The map analysis of fracture surface of copper side.

19

Iranian Journal of Materials Science and Engineering, Vol. 4, Numbers 3 and 4, Summer and Autumn 2007

The Mo and MgAl2O4 particles have marked in Fig. 12. As illustrated in Fig. 12, Mo continuous network have slashed in some points. In other words, crack has progressed in these areas without any path change and has ruptured Mo particles. The fracture surface on spinel areas in some points is in higher level and in some points is in lower level than Mo areas. Propagation path has, therefore, changed in these areas and have rounded spinel particles. The unchanged shape of spinel particles in Fig. 9 verifies this deduction. As a result, it can be said that the spinel particles serve as obstacles on crack propagation and crack growth has been delayed in these points. To sum up, it can be concluded that the crack has nucleated in glass phase. Crack propagation has continued in glass phase and by striking to spinel particles its path has changed. Following, crack was rounded spinel particles without any fracture in them. By clash to Mo particles and slight plastic deformation, crack has slashed the Mo particles and has ruptured the joint. So, spinel particles have functioned as hindrances to crack propagation and have caused an increase in strength. Nevertheless, formation of more spinel phase causes a decrease in the percentage of glassy phase. Some pores in interface will not, consequently, filled with glassy phase and can result to strength decrease. Also, homogeneous distribution of Mo particles in glass phase can delay the crack propagation owing to their plastic deformations. 4. CONCLUSIONS 1. By reducing the percentage of glass phase to cease the distinct glass phase formation, the strength of alumina- copper joint has increased. However, excess reduction of glass phase can be detrimental to joint strength due to unfilled pores and low mechanical connection with Mo continuous network. 2. The crack propagation path by reducing the percentage of glass phase and introducing more spinel phase has changed. In other words, instead of propagation through the brittle glass phase, crack propagation has been hindered by encountering to spinel particles, and without any fracture in spinel particles, has rounded them. Following, by

20

slashing the Mo particles, crack propagation has caused the joint rupture. 3. Spinel particles in metalized layer can serve as obstacles on crack propagation path and can increase the strength of the joint. 4. The plastic deformation of Mo particles before fracture has slowed down the crack propagation; homogenous distribution of Mo particles in metalized layer can, hence, increase the peel strength of alumina-copper joint. ACKNOWLEDGEMENT 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, and Mr. E. Nazari. REFERENCES 1. Taira Okamoto, "Interfacial Structure of Metal-Ceramic Joints”, ISI International Journal, 1990, Vol. 30, No. 12, pp: 1033 – 1040. 2. J. Intrater, “Review of Some Processes for Ceramic-to-Metal Joining”, Material & Manufacturing Processes, 1993, 8(3), pp: 353-373. 3. Sudipta Mandal, Ashok Kumar Ray, Ajoy Kumar Ray, “Correlation between the mechanical properties and the microstructural behavior of Al2O3 – (Ag-CuTi) brazed joints”, Materials Science & Engineering A, 2004, 383, pp: 235-244. 4. R. M. Do Nascimento, A. E. Martinelli, A. J. A.Buschineli, “Review Article: Recent advances in metal-ceramic brazing”, Ceramica, 2003, 49, pp: 178-198. 5. M. M. Schwartz, “Ceramic Joining “, 1990, ASM International, pp: 87-90. 6. Charles A. Lewinsohn, Mrityunjay singh, Rouald Loehman, “Advances in joining of ceramics”, 2003, American Ceramic Society. pp: 83-84. 7. Michikazu Kinsho, Yoshio Saito, Zenzaburo Kabeya, Keisuke Tajiri, Tomaru Nakamura, Kazuhiko Abe, Taketoshi Nagayama, Daiji Nishizawa, Norio Ogiwara, "Development of alumina ceramics vacuum duct for the 3 GeV-RCS of the J-PARC project ", Vacuum,

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

8.

9.

10.

11.

2004, 73, pp: 187–193 . Lidija I.Belic, ALenka S. Kosmos,” Synthesis reactions and thermodynamics of compound formation during MoMn layer sintering”, Fizika A, 1995, 4(2), pp: 125129. A. j. Moorhead and W. H. Elliott, “Brazing of Ceramic and Ceramic to Metal Joints “, ASM Handbook, Vol .6, 1996. K. Suganuma, Y. Miyamoto, M. Koizumi, "Joining of Ceramics and Metals ", Mater. Sci., 1988, 18, pp: 47-73 . Adhesives –Determination of Peel resistance of flexible- bonded- to- rigid specimen assemblies by 180° peel test, ISO 8510-2.

21