- Email: [email protected]
Ni plated Cu substrates at ambient temperature
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 179–186
journal homepage: www.elsevier.com/locate/jmatprotec
Investigation of ultrasonic copper wire wedge bonding on Au/Ni plated Cu substrates at ambient temperature Yanhong Tian a,∗ , Chunqing Wang a , Ivan Lum b , M. Mayer b , J.P. Jung c , Y. Zhou a,b a b c
State Key Laboratory of Advanced Welding Production Technology, Harbin Institute of Technology, Harbin 150001, China Microjoining Lab, Center for Advanced Materials Joining, University of Waterloo, Waterloo, Canada N2L 3G1 Microjoining Lab, University of Seoul, Seoul, Republic of Korea
a r t i c l e
i n f o
a b s t r a c t
Article history:
Copper wire is attracting more and more attention in wire bonding technology due to its
Received 24 December 2006
advantages in comparison with gold or aluminum wire. This paper presents an achievement
Received in revised form
of ultrasonic wedge bonding with 25 m copper wire on Au/Ni plated Cu substrate at ambi-
6 December 2007
ent temperature. A detailed investigation from the aspects of process optimization, bonding
Accepted 23 December 2007
mechanism, interdiffusion, ultrasonic effects on microstructure and microhardness of the bonding materials were performed. The results show that it is possible to produce strong copper wire wedge bonds at room temperature, and the thinning of the Au layer was found
Keywords:
directly below the center of the bonding tool with the bonding power increasing. Interdif-
Copper wire
fusion between copper wire and Au metallization during the wedge bonding at ambient
Ultrasonic wedge bonding
temperature was assumed negligible. The wedge bonding was achieved by wear action
Design of experiment (DOE)
induced by ultrasonic vibration. The ultrasonic power did contribute to enhancing defor-
Wear action
mation of the copper wire due to ultrasonic softening effect which was then followed by the
Ultrasonic softening
strain hardening of the copper wedge bond, and the dynamic recovery or recrystallization
Recrystallization
of the copper wire caused by ultrasonic vibration during wedge bonding was also found. © 2008 Elsevier B.V. All rights reserved.
1.
Introduction
Copper wire bonding is an alternative chip interconnection technology with promising cost savings compared to gold wire bonding and better electrical performance compared to aluminum wire (Harman, 1997). There are lots of studies on thermosonic gold or copper ball bonding and ultrasonic aluminum wedge bonding (Ho et al., 2003; Harman and Albers, 1977; Krzanowski et al., 1990; Takahashi et al., 1996; Langenecker, 1966; Lum et al., 2005, 2006; Murali et al., 2003; Li et al., 2006). However, there is a lack of understanding on the ultrasonic copper wire wedge bonding process. Ultrasonic wedge bonding utilizes a normal bond force simultaneously with ultrasonic energy to form the first and second bonds at
∗
ambient temperatures and is a preferred method in interconnecting power devices. Ultrasonic wire bonding is generally accepted to be a solid state joining process which is supported by various evidences such as bonds made at liquid nitrogen temperatures (Harman and Albers, 1977) and studies of the bond interface with transmission electron microscopy (Krzanowski et al., 1990). A major requirement to form a metallurgical bond is a relatively contaminant free surface. Without occurrence of melting in the wire-bonding process other methods of contaminant dispersal are required in order to facilitate bonding. Deformation is the main mechanism responsible for the contaminant dispersal required for bond formation in thermocompression (using heat and pressure only) wire bonding. The deforma-
Corresponding author. Tel.: +86 451 86418359; fax: +86 451 86416186. E-mail addresses: [email protected], [email protected] (Y. Tian). 0924-0136/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.jmatprotec.2007.12.134
180
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 179–186
tion mechanism observed in thermocompression bonding is similarly observed in pressure welding, in which applied pressure and the subsequent deformation breaks up the oxide layer (Takahashi et al., 1996). In ultrasonic wire bonding, ultrasonic energy is used in addition to pressure. When a metal is irradiated with ultrasonic energy, the yield stress decreases, and this is known as the reversible ultrasonic softening effect (Langenecker, 1966). Recently, Lum et al. proposed the transition from micro-slip to gross slide with ultrasonic power increasing based on the micro-slip theory to elucidate the ball and wedge bonding mechanism (Lum et al., 2005, 2006). Murali et al., 2003 reported their un-annealed copper wire after ball bonding showed the lowest hardness in the HAZ zone, which was caused by the recrystallization and grain growth from the FAB formation process. They concluded that the highest hardness in the Cu ball bond came from the strain hardening induced by ultrasonic power. Li et al., 2006 found the atomic diffusion between Au ball bond and Al pad at a high level ultrasonic frequency (1.5 MHz) when ultrasonic power is 1.75 W and bonding temperature 200 ◦ C, and the thickness of atomic diffusion layer is about 500 nm. This paper will present a detailed investigation on the ultrasonic wedge bonding of copper wire at ambient temperature, and many aspects of the copper wire wedge bonding including process optimization, bonding mechanism, microstructure, and microhardness of the copper wedge bond and interdiffusion of the bonding materials were studied.
2.
Experimental procedure
Copper wire bonding was performed at room temperature on Au/Ni plated Cu substrate with the 3 m thickness of Au, 7 m Ni and 23 m Cu. The 25 m copper wire used is provided by MK Electron Co. Ltd. with 99.99% purity. The bonding machine used was a semi-automatically 4523A Digital K&S wedge bonder with a frequency of 65 KHz. Fig. 1 illustrates the Cu wire wedge bonding process. Bond growth and joint strength are related with processing parameters and ultrasonic conditions such as ultrasonic vibration accuracy and speed, ultrasonic power, bonding force, bonding time, tail breaking force, surface state, etc. These parameters should be optimized to get good bond quality at ambient temperature. Design of experiment (DOE) was applied in this study, and a 20-run central composite design based on response surface methodology was used. There are three
Fig. 1 – Schematic drawing of the Cu wire wedge bonding process.
Fig. 2 – Three types of the first bond outcomes during wedge bonding (a) lift-off (b) sticking, and (c) the first bond cut caused by excessive deformation.
factors in the DOE, ultrasonic power, bonding force, and bonding time. For each factor, three levels were chosen. There are 20 runs in total, and for each set of parametric conditions 20 bonds were made. After wedge bonding, pull testing was performed on DAGE 4000 to get the pull force of the bond. Pull force was the response for this design. The experimental design and data analysis were done on MiniTAB statistical software. The cross-section samples were prepared and chemically etched. An etch solution (containing 2 g Na2 Cr2 O7 + 4 ml saturated NaCl solution + 10 ml concentrated H2 SO4 + 100 ml H2 O) was swabbed onto the cross-section surface for 12 s using a cotton ball to reveal the microstructure. The cross-section samples were observed using SEM, and energy dispersive Xray (EDX) was used to study the chemical composition. Vickers microhardness testing was conducted on the cross-section of the copper wedge bonds at various locations. The method employs an indentation measurement by using a 136◦ diamond pyramid indenter and 5 gf load and was applied for 15 s.
3.
Results and discussion
3.1. Effects of process parameters on the pull force of wedge bonds During the wedge bonding of the copper wire, three types of bonding outcomes were obtained: lift-off caused by weak bonding, sticking, and wedge bond cut caused by excessive deformation, as illustrated in Fig. 2. Lifted off bonds would occur because the frictional force acting at the wire/wire feed hole during the following looping step would be greater than the strength of the wedge bond and during the following looping step would lift the bond off the substrate. On the other hand, if the bond was sufficiently strong, the wedge bond would stick on the substrate during the following looping step. The completed wire bond (sticking bond in Fig. 2b) would subsequently be pull tested with a DAGE 4000 pull tester, and then pull force and failure modes are determined. Three common failure modes in this study are: 1) Interfacial break (bond lifting off from the surface of the metallization). 2) Neck break (wire break at the neck of the bond).
181
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 179–186
Table 1 – Parameter levels in the central composite design Variables
Level of variables
Power (mW) Time (ms) Force (gf)
Fig. 3 – Plot of pull test results illustrating the different regions of failure modes and resulting pull force with different ultrasonic powers.
3) Bond break (when the bond was deformed excessively). The failure mode may give insight to the wire bondability and the bond strength. Wire breaks at the neck position of the bond are the preferred mode in this experiment because a high wire load with a wire break indicates good bonding between the wire and the Au/Ni/Cu metallization. Bond breaks are not preferred because the bonds deformed excessively under high bonding power and force. Fig. 3 shows the pull test results of the first wedge bonds formed by the different ultrasonic powers when the bonding time and force was fixed (30 ms/40 gf, 1 gf = 9.8 mN). Fig. 3 also illustrates the different regions of failure modes. With 65 mW and less ultrasonic bonding power (labeled lift-off region), lift-off bonds would occur since the amount of bonding was minimal. Bond sticking would occur with ultrasonic bonding power higher than 65 mW. With increased bonding power up to 390 mW, the percentage of interfacial break decreased and more neck breaks occurred. It can be seen that the wedge bonds with the neck break failure modes yields the highest pull forces during pulling test. Fig. 4 shows pull force of the first wedge bonds formed by various bonding forces when the ultrasonic power and bonding time was fixed (260 mW/30 ms). Here, a pull force of larger
Fig. 4 – Pull force of wedge bonds formed by various bond forces.
−␣
−1
0
+1
+␣
150 14 32
195 20 35
260 30 40
325 40 45
370 46 48
than 20 g is chosen as a suitable condition for the bonding force process window. This condition is fulfilled in the window between 35 and 50 g. Below these bonding forces, contact pressure is insufficient, resulting in a loose contact. At other extreme, excessive bonding forces are detrimental to the interfacial motion of wire and pad. Consequently, the resulting bond has a lower pull force.
3.2.
DOE process optimization of the wedge bonds
Design of experiment is a quick and cost-effective method to understand and optimize any manufacturing processes (Raymond and Douglas, 2002). It is a direct replacement of ‘one variable-at-a-time’ approach of experimentation, where experimenters vary only one variable at a time, keeping all other variables in the experiment fixed. In order to optimize the process parameters of both the first bond and second bond, a 20-run central composite design was used in this study. Based on the aforementioned experiment results, three levels of the factors including ultrasonic power, bonding force, and bonding time were determined. Table 1 shows the parameter levels for the central composite design. In this paper, pull force of the wedge bonds was used as a response or quality characteristic, that is also the output we want to mea-
Table 2 – Details of the ultrasonic power, bonding time, bonding force, bonding responses (pull force), and standard deviation (StdEv) for each run of DOE for the first bonds Run no. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Power (mW) 325 260 325 325 260 195 195 195 325 195 260 260 260 150 260 260 260 370 260 260
Time (ms) 20 30 40 40 30 40 20 20 20 40 30 30 30 30 14 30 46 30 30 30
Force (gf)
Pull force (gf)
StdEv
35 40 35 45 40 35 45 35 45 45 40 40 40 40 40 32 40 40 48 40
22.8 22.3 22.0 19.4 20.6 20.9 18.6 18.9 19.0 20.6 20.7 22.2 21.6 20.9 19.1 21.1 19.0 19.6 19.6 21.1
2.104 1.905 2.36 1.172 1.993 1.743 2.27 2.375 2.14 1.954 0.948 0.964 1.199 1.973 2.181 2.510 1.443 1.468 1.007 2.876
182
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 179–186
Table 3 – Details of the ultrasonic power, bonding time, bonding force, bonding responses (pull force), and standard deviation (StdEv) for each run of DOE for the second bonds Run no. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Power (mW) 325 260 325 325 260 195 195 195 325 195 260 260 260 150 260 260 260 370 260 260
Time (ms)
Force (gf)
20 30 40 40 30 40 20 20 20 40 30 30 30 30 14 30 46 30 30 30
35 40 35 45 40 35 45 35 45 45 40 40 40 40 40 32 40 40 48 40
Pull force (gf) 16.1 18.0 15.4 20.2 19.8 10.0 10.1 5.9 20.4 9.1 19.2 19.5 16.4 7.4 9.8 13.2 17.2 20.5 17.9 19.8
StdEv 3.75 3.44 4.46 3.02 2.14 3.05 3.495 2.424 1.62 3.17 3.41 3.66 2.47 3.95 3.55 6.03 5.71 2.86 2.77 1.32
sure during the experiment. Tables 2 and 3 show details of the ultrasonic power, bonding time, bonding force, bonding responses (pull force), and standard deviation (StdEv) for each run for both the first and second bonds. It can be found that for the copper wire used in this study, the pull forces of 20 gf or greater can be obtained for both the first and second bonds. The higher pull force and lower standard deviation of the first bonds can be achieved compared with the second bonds. Fig. 5 shows the contour plots of the pull force when the bonding time is 30 ms. It can be found that the highest pull force of the first bond was achieved with high power and low force. However, for the second bond, for the highest pull force, both high power and high force were required. This might be because of the tail formation with the wire clamp which closes and pulls the wire to break it at the heel of the second bond that requires more force and power. However, there is no pulling force of the clamp on the first bond, as shown in Fig. 1. According to the contour plots, the optimized ultrasonic power, bonding force, and bonding time for both the first bond and the second bond could be achieved, which are 260 mW/35 gf/30 ms and 325 mW/40 gf/30 ms, respectively. Fig. 6(a) shows the wedge bonds obtained with the optimized processing parameters when the bonding time is 30 ms with first bonds at a power of 260 mW and bonding force of 35 gf, and the second bonds at a power of 325 mW and bonding force of 40 gf. Fig. 6(b) shows the first bonds obtained with power 370 mW, force 40 gf, and time 30 ms. It could be found that with the increase of ultrasonic power, the deformation of the wedge bond increased, and excessive deformation of the first bonds occurred when a higher ultrasonic power was applied.
Fig. 5 – Contour plots of pull force vs. bonding power and bonding force when bonding time is 30 ms. (a) First bond and (b) second bond.
3.3.
Microstructure and hardness of the wedge bond
The possibility of using Cu wires bonded to Au/Ni plated Cu substrate has led to interest in the reliability of this metallurgical system. Fig. 7 shows the cross-sections of second bonds when the bonding force is 35 gf, bonding time is 30 ms and ultrasonic power is 260 and 370 mW, respectively. With the bonding power increasing, the thickness of the gold layer at the center of the Au/Ni metallization decreased, as shown in Fig. 7b, this will be discussed in detail at the following section. It seems that the ultrasonic power contributed to increased deformation of the copper wire because higher ultrasonic power made the wire softer due to ultrasonic softening. Fig. 8 shows two lines scan from the same cross-sections of Fig. 7a and b, which indicates that there was no obvious interdif-
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 179–186
183
Fig. 8 – Results of line scan from the same cross-sections of Fig. 5: (a) line scan from line 1 of Fig. 5a and (b) line scan from line 2 of Fig. 5b. Fig. 6 – Micrographs of the wedge bonds when the bonding time is 30 ms (a) the wedge bonds obtained with the optimized processing parameters, the first bonds (260 mW, 35 gf); the second bonds (325 mW, 40 gf); (b) the first bonds with high ultrasonic power shows excessive deformation (370 mW, 35 gf).
fusion between Cu and Au at these two bonding interfaces, however, a little amount of Au elements inside the Cu wire was found, which might be caused by the wear action and mechanical mixing.
Fig. 9 shows fractured surfaces after shear testing. Both of the fracture surfaces showed dimples which indicated desirable ductile bonded joints. The composition of the fracture surface was Au and Cu, which showed that the fracture occurred along the bonded interface between the Cu and Au layer. For 325 mW bonds, top layer of Au/Ni metallization lifted off somewhere and the Cu substrate was exposed during shearing, which could be found from EDX result of point A. Point B and point C show good bonding and mixing of the ele-
Fig. 7 – Cross-sections of second bonds with different bonding power showing thinning of the Au layer occurred with ultrasonic power increasing. (a) 260 mW, 35 gf, 30 ms and (b) 370 mW, 35 gf, 30 ms.
184
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 179–186
Fig. 9 – Fractured surface of the wedge bond after shear test (a) and (b) 325 mw; (c) and (d) 370 mw.
ments between wire and metallization, as shown in Fig. 9b. For the bonds with increased power of 370 mW, a small white particle was found at the fractured surface, as shown in Fig. 9c and d. The EDX result of point D shows that the particle consists of Cu and Au. There are three intermetallic phases (Cu3 Au, CuAu, CuAu3 ) in the binary Cu–Au phase diagram when the temperature is beyond 200 ◦ C. According to the literature, Cu moves rapidly through the Au film by boundary diffusion at temperatures of 100–300 ◦ C within approximately 1 h, and the diffusion coefficient for Cu in Au is D = 1.64 × 10−20 cm2 /s at 200 ◦ C (Hall and Morabito, 1978). This diffusion is faster than that of Au in Cu because the Cu atom is smaller than the Au atom. It was found from some ultrasonic wire bonding investigations that temperature rise at the bond interface was between 80 and 300 ◦ C (Ho, 2004). According to the Fick’s diffusion law, the thermal interdiffusion distance can be obtained from following equation: X2 = Dt, where t is the interdiffusion time, which is assumed as the bonding time 50 ms here. As a result of this, the thermal diffusion distance is not more than 0.3 A◦ . It was proposed that the interdiffusion between Al wire and Ni metallization was enhanced by ultrasonic vibration, and atomic diffusion between Au bond and Al metallization at a high level ultrasonic frequency (1.5 MHz) and bonding temperature 200 ◦ C was also found (Li et al., 2006). In this paper, the wedge bonding of the Cu wire was achieved at ambient temperature, and the interdiffusion between Cu and Au was not found from Fig. 8. Therefore it is concluded that the Cu diffusion into the Au during the wedge bonding at ambient temperature was negligible, and the formation of IMCs is not expected.
The above results confirmed that the mixing of the Cu and Au at the interface region and the achievement of the wedge bonding was caused by wearing action. In ultrasonic wedge bonding, the amplitude of the bonding tip oscillation is proportional to the applied ultrasonic power and the relative motion experienced when the wire is sliding, will lead to wear of material (or contaminant) according to an equation developed for contacting surfaces in relative motion (Peterson and Winer, 1980):
t=d
H 1 K PV
(1)
where t is the time required, d is the depth of material worn, P is the mean or nominal pressure, H is the hardness of the material, K is the wear coefficient constant, and V is the sliding velocity. This wear of material is termed fretting when small amplitude oscillations are involved. A sufficient removal of the contaminant layer is required for bonding to occur between the underlying metal surfaces. Bonding at room temperature is related to a wear mechanism induced by the ultrasonic vibrations. During bonding, the ultrasonic stick–slip friction causes friction power to be delivered at the interface. This power is partly transformed to mechanical wear. The wearing action breaks up contaminant and oxide layers allowing for areas of fresh metal of the opposing bonding partners to contact and bond to each other. From Fig. 9(a) and (c), the sheared fracture surface (footprint) showed the typical ellipse shape of wedge bonding. It is the wear caused by the relative motion that will allow intimate metal–metal contact and promotes subsequent bonding.
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 179–186
185
Fig. 10 – Microstructures of bonded joints when power is 325 mW, bonding force 35 gf, and time 50 ms (a) copper wire showing elongated grains in the middle (b) magnified picture showing the small equiaxed grains.
Fig. 11 – Microhardness test on the wedge bond when power is 260 mW, time 50 ms, and force 35 gf. (a) The first bond and (b) the second bond.
These regimes of relative motion combined with the stress distribution at the bonding interface caused by the bonding tool that account for the bonded footprint morphology. The rate and uniformity of the wear at the interface depends on the stress field amplitudes and uniformity at the interface, respectively. Because the stress is lowest right at the periphery of the contact, and is highest at the bonding interface directly below the tool. The thinning of the Au layer observed in Fig. 7b can be explained by the larger interfacial peak stresses right below the wedge tool where the thinning is observed when the higher ultrasonic power is used. Fig. 10 gives the microstructure of the bonded joint where ultrasonic power, bonding force, and time are 325 mW, 35 gf and 50 ms, respectively. The microstructures of the copper wire in the left upper part of Fig. 10a are elongated grains, which were produced during the drawing process of the wires when manufacturing. After wedge bonding, small equiaxed grains were found at the bonding interface region, as shown in Fig. 10b. As we discussed above, the temperature rising during room temperature wedge bonding process was not high enough to cause any microstructure change and even recovery of copper wire, thus it can be concluded that the recovery or recrystallization found in this study was due to the ultrasonic vibration. The horizontal ultrasonic vibration and the normal force of the wedge induced compressive stress and shear stress at the interfacial region, which resulted in the slipping and refining the grains the wedge bond. The slipping of the copper wire and refining grains provided a foundation of the further plastic deformation. The results were consistent with
the others studies in which the dynamic recovery or recrystallization of the aluminum wire was found during ultrasonic wedge bonding (Geissler et al., 2006; Krzanowski, 1990). Micrographs of microhardness test results are given in Fig. 11. The test was performed along the centerline of the copper wire. Five points for the wire part and the bonded joint part were tested. It could be found that the hardness at the deformed bonded joint part is a little higher than the wire part, which was probably caused by the strain hardening of the wedge bond. It was believed that the ultrasonic energy enhances the plastic deformation of the bonding wire due to the fact that the dislocations inside the wire absorbs the acoustic energy selectively, and the dislocations were activated from their anchoring locations, which makes the deformation of the bonding wire easier. After the acoustic energy was removed, the new defects made the bonding wire harder.
4.
Conclusions
(1) Copper wire bonding on an Au/Ni plated Cu substrate at ambient temperature was achieved, and the bonding parameters for both first bond and second bonds were optimized by design of experiment. To get strong pull force of the wedge bond, higher power and force were required for the second bond than for the first bond. (2) Cross-section analysis showed a continuous connection between the Cu wire and Au metallization when the
186
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 179–186
appropriate bonding parameters were chosen. With the increase of the ultrasonic power, the thickness of the gold layer at the center of the metallization was found. This is because the interfacial peak stresses occurred right below the wedge tool where the thinning is observed. (3) The interdiffusion between Cu wire and Au layer at the bonding interface during the wedge bonding at ambient temperature was assumed negligible. The achievement of copper wedge bonding was proposed to be the wear action and mechanical mixing induced by the ultrasonic vibration, and the ultrasonic power contributes to increasing deformation of the copper wire due to ultrasonic softening which was then followed by the strain hardening of the copper wedge bond, and the dynamic recovery or recrystallization of the copper wire caused by ultrasonic vibration during wedge bonding was also found. (4) Since the Au metallization layer has a good bondability, room temperature wedge bonding of the copper wire is feasible. However, if the bare Cu or Al substrates are used, it becomes very diffucult. In the future work, different kinds of the metallization layers should be investigated, and oxidation protection, coated copper wire, as well as the copper wire with different elongation should also be studied further.
Acknowledgments The research work was supported by National Natural Science Foundation of China under grant No. 50705021/E052104, MK Electron Ltd. of South Korea and the Development Program for Outstanding Young Teachers in HIT (Project No.: HIT. 2006: 01504489).
references
Geissler, U., Martin, S.-R., Lang, K.-D., Reichl, H., 2006. Investigation of microstructural processes during ultrasonic wedge/wedge bonding of AlSi1 wires. J. Electron. Mater. 35, 173–180.
Hall, P.M., Morabito, J.M., 1978. Diffusion problems in microelectronics packages. Thin Solid Films 53, 175–182. Harman, G.G., 1997. Wire Bonding in Microelectronics—Materials, Processes, Reliability, and Yield, 2nd edition. McGraw Hill, New York, NY. Harman, G.G., Albers, J., 1977. Ultrasonic welding mechanism as applied to aluminum-and gold-wire bonding in microelectronics. IEEE Trans. Parts, Hybr., Pack. 13, 406–411. Ho, J.R., 2004. Thin film thermal sensor for real time measurement of contact temperature during ultrasonic wire bonding process. Sens. Actuator A: Phys. 111, 188–195. Ho, H.M., Lam, W., Stoukatch, S., Ratchev, P., Vath III, C.J., Beyne, E., 2003. Direct gold and copper wires bonding on copper. Microelectron. Reliab. 43, 913–923. Krzanowski, J.E., 1990. Transmission electron microscopy study of ultrasonic wire bonding. IEEE Trans. Compon., Hybr., Manuf. Technol. 13, 361–368. Krzanowski, J.E., Razon, E., Hmiel, A.F., 1990. The effect of thin film structure and properties on gold ball bonding. J. Electron. Mater. 19, 919–924. Langenecker, B., 1966. Effects of ultrasound on deformation characteristics of metals. IEEE Trans. Son. Ultrason. 13, 1–8. Li, J.H., Wang, F., Han, L., Duan, J., Zhong, J., 2006. Atomic diffusion properties in wire bonding. Trans. Nonferrous Met. Soc. China 16, 463–466. Lum, I., Jung, J.P., Zhou, Y., 2005. Bonding mechanism in ultrasonic gold ball bonds on copper substrate. Metall. Mater. Trans. 36A, 1279–1286. Lum, I., Mayer, M., Zhou, Y., 2006. Footprint study of ultrasonic wedge-bonding with aluminum wire on copper substrate. J. Electron. Mater. 35, 422–433. Murali, S., Srikanth, N., Vath III, C.J., 2003. Grains, deformation substructures, and slip bands observed in thermosonic copper ball bonding. Mater. Charact. 50, 39–50. Peterson, M.B., Winer, W.O., 1980. Wear Theory and Mechanism from Wear Control Handbook. ASME United Engineering Center. Raymond, M., Douglas, C., 2002. Response Surface Methodology: Process and Product Optimization Using Designed Experiments, 2nd edition. John Wiley & Sons, Inc., New York. Takahashi, Y., Shibamoto, S., Inoue, K., 1996. Numerical analysis of the interfacial contact. Process in wire thermocompression bonding. IEEE Trans. Compon. Hybr. Manuf. Technol. Part A 19, 351–356.
journal homepage: www.elsevier.com/locate/jmatprotec
Investigation of ultrasonic copper wire wedge bonding on Au/Ni plated Cu substrates at ambient temperature Yanhong Tian a,∗ , Chunqing Wang a , Ivan Lum b , M. Mayer b , J.P. Jung c , Y. Zhou a,b a b c
State Key Laboratory of Advanced Welding Production Technology, Harbin Institute of Technology, Harbin 150001, China Microjoining Lab, Center for Advanced Materials Joining, University of Waterloo, Waterloo, Canada N2L 3G1 Microjoining Lab, University of Seoul, Seoul, Republic of Korea
a r t i c l e
i n f o
a b s t r a c t
Article history:
Copper wire is attracting more and more attention in wire bonding technology due to its
Received 24 December 2006
advantages in comparison with gold or aluminum wire. This paper presents an achievement
Received in revised form
of ultrasonic wedge bonding with 25 m copper wire on Au/Ni plated Cu substrate at ambi-
6 December 2007
ent temperature. A detailed investigation from the aspects of process optimization, bonding
Accepted 23 December 2007
mechanism, interdiffusion, ultrasonic effects on microstructure and microhardness of the bonding materials were performed. The results show that it is possible to produce strong copper wire wedge bonds at room temperature, and the thinning of the Au layer was found
Keywords:
directly below the center of the bonding tool with the bonding power increasing. Interdif-
Copper wire
fusion between copper wire and Au metallization during the wedge bonding at ambient
Ultrasonic wedge bonding
temperature was assumed negligible. The wedge bonding was achieved by wear action
Design of experiment (DOE)
induced by ultrasonic vibration. The ultrasonic power did contribute to enhancing defor-
Wear action
mation of the copper wire due to ultrasonic softening effect which was then followed by the
Ultrasonic softening
strain hardening of the copper wedge bond, and the dynamic recovery or recrystallization
Recrystallization
of the copper wire caused by ultrasonic vibration during wedge bonding was also found. © 2008 Elsevier B.V. All rights reserved.
1.
Introduction
Copper wire bonding is an alternative chip interconnection technology with promising cost savings compared to gold wire bonding and better electrical performance compared to aluminum wire (Harman, 1997). There are lots of studies on thermosonic gold or copper ball bonding and ultrasonic aluminum wedge bonding (Ho et al., 2003; Harman and Albers, 1977; Krzanowski et al., 1990; Takahashi et al., 1996; Langenecker, 1966; Lum et al., 2005, 2006; Murali et al., 2003; Li et al., 2006). However, there is a lack of understanding on the ultrasonic copper wire wedge bonding process. Ultrasonic wedge bonding utilizes a normal bond force simultaneously with ultrasonic energy to form the first and second bonds at
∗
ambient temperatures and is a preferred method in interconnecting power devices. Ultrasonic wire bonding is generally accepted to be a solid state joining process which is supported by various evidences such as bonds made at liquid nitrogen temperatures (Harman and Albers, 1977) and studies of the bond interface with transmission electron microscopy (Krzanowski et al., 1990). A major requirement to form a metallurgical bond is a relatively contaminant free surface. Without occurrence of melting in the wire-bonding process other methods of contaminant dispersal are required in order to facilitate bonding. Deformation is the main mechanism responsible for the contaminant dispersal required for bond formation in thermocompression (using heat and pressure only) wire bonding. The deforma-
Corresponding author. Tel.: +86 451 86418359; fax: +86 451 86416186. E-mail addresses: [email protected], [email protected] (Y. Tian). 0924-0136/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.jmatprotec.2007.12.134
180
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 179–186
tion mechanism observed in thermocompression bonding is similarly observed in pressure welding, in which applied pressure and the subsequent deformation breaks up the oxide layer (Takahashi et al., 1996). In ultrasonic wire bonding, ultrasonic energy is used in addition to pressure. When a metal is irradiated with ultrasonic energy, the yield stress decreases, and this is known as the reversible ultrasonic softening effect (Langenecker, 1966). Recently, Lum et al. proposed the transition from micro-slip to gross slide with ultrasonic power increasing based on the micro-slip theory to elucidate the ball and wedge bonding mechanism (Lum et al., 2005, 2006). Murali et al., 2003 reported their un-annealed copper wire after ball bonding showed the lowest hardness in the HAZ zone, which was caused by the recrystallization and grain growth from the FAB formation process. They concluded that the highest hardness in the Cu ball bond came from the strain hardening induced by ultrasonic power. Li et al., 2006 found the atomic diffusion between Au ball bond and Al pad at a high level ultrasonic frequency (1.5 MHz) when ultrasonic power is 1.75 W and bonding temperature 200 ◦ C, and the thickness of atomic diffusion layer is about 500 nm. This paper will present a detailed investigation on the ultrasonic wedge bonding of copper wire at ambient temperature, and many aspects of the copper wire wedge bonding including process optimization, bonding mechanism, microstructure, and microhardness of the copper wedge bond and interdiffusion of the bonding materials were studied.
2.
Experimental procedure
Copper wire bonding was performed at room temperature on Au/Ni plated Cu substrate with the 3 m thickness of Au, 7 m Ni and 23 m Cu. The 25 m copper wire used is provided by MK Electron Co. Ltd. with 99.99% purity. The bonding machine used was a semi-automatically 4523A Digital K&S wedge bonder with a frequency of 65 KHz. Fig. 1 illustrates the Cu wire wedge bonding process. Bond growth and joint strength are related with processing parameters and ultrasonic conditions such as ultrasonic vibration accuracy and speed, ultrasonic power, bonding force, bonding time, tail breaking force, surface state, etc. These parameters should be optimized to get good bond quality at ambient temperature. Design of experiment (DOE) was applied in this study, and a 20-run central composite design based on response surface methodology was used. There are three
Fig. 1 – Schematic drawing of the Cu wire wedge bonding process.
Fig. 2 – Three types of the first bond outcomes during wedge bonding (a) lift-off (b) sticking, and (c) the first bond cut caused by excessive deformation.
factors in the DOE, ultrasonic power, bonding force, and bonding time. For each factor, three levels were chosen. There are 20 runs in total, and for each set of parametric conditions 20 bonds were made. After wedge bonding, pull testing was performed on DAGE 4000 to get the pull force of the bond. Pull force was the response for this design. The experimental design and data analysis were done on MiniTAB statistical software. The cross-section samples were prepared and chemically etched. An etch solution (containing 2 g Na2 Cr2 O7 + 4 ml saturated NaCl solution + 10 ml concentrated H2 SO4 + 100 ml H2 O) was swabbed onto the cross-section surface for 12 s using a cotton ball to reveal the microstructure. The cross-section samples were observed using SEM, and energy dispersive Xray (EDX) was used to study the chemical composition. Vickers microhardness testing was conducted on the cross-section of the copper wedge bonds at various locations. The method employs an indentation measurement by using a 136◦ diamond pyramid indenter and 5 gf load and was applied for 15 s.
3.
Results and discussion
3.1. Effects of process parameters on the pull force of wedge bonds During the wedge bonding of the copper wire, three types of bonding outcomes were obtained: lift-off caused by weak bonding, sticking, and wedge bond cut caused by excessive deformation, as illustrated in Fig. 2. Lifted off bonds would occur because the frictional force acting at the wire/wire feed hole during the following looping step would be greater than the strength of the wedge bond and during the following looping step would lift the bond off the substrate. On the other hand, if the bond was sufficiently strong, the wedge bond would stick on the substrate during the following looping step. The completed wire bond (sticking bond in Fig. 2b) would subsequently be pull tested with a DAGE 4000 pull tester, and then pull force and failure modes are determined. Three common failure modes in this study are: 1) Interfacial break (bond lifting off from the surface of the metallization). 2) Neck break (wire break at the neck of the bond).
181
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 179–186
Table 1 – Parameter levels in the central composite design Variables
Level of variables
Power (mW) Time (ms) Force (gf)
Fig. 3 – Plot of pull test results illustrating the different regions of failure modes and resulting pull force with different ultrasonic powers.
3) Bond break (when the bond was deformed excessively). The failure mode may give insight to the wire bondability and the bond strength. Wire breaks at the neck position of the bond are the preferred mode in this experiment because a high wire load with a wire break indicates good bonding between the wire and the Au/Ni/Cu metallization. Bond breaks are not preferred because the bonds deformed excessively under high bonding power and force. Fig. 3 shows the pull test results of the first wedge bonds formed by the different ultrasonic powers when the bonding time and force was fixed (30 ms/40 gf, 1 gf = 9.8 mN). Fig. 3 also illustrates the different regions of failure modes. With 65 mW and less ultrasonic bonding power (labeled lift-off region), lift-off bonds would occur since the amount of bonding was minimal. Bond sticking would occur with ultrasonic bonding power higher than 65 mW. With increased bonding power up to 390 mW, the percentage of interfacial break decreased and more neck breaks occurred. It can be seen that the wedge bonds with the neck break failure modes yields the highest pull forces during pulling test. Fig. 4 shows pull force of the first wedge bonds formed by various bonding forces when the ultrasonic power and bonding time was fixed (260 mW/30 ms). Here, a pull force of larger
Fig. 4 – Pull force of wedge bonds formed by various bond forces.
−␣
−1
0
+1
+␣
150 14 32
195 20 35
260 30 40
325 40 45
370 46 48
than 20 g is chosen as a suitable condition for the bonding force process window. This condition is fulfilled in the window between 35 and 50 g. Below these bonding forces, contact pressure is insufficient, resulting in a loose contact. At other extreme, excessive bonding forces are detrimental to the interfacial motion of wire and pad. Consequently, the resulting bond has a lower pull force.
3.2.
DOE process optimization of the wedge bonds
Design of experiment is a quick and cost-effective method to understand and optimize any manufacturing processes (Raymond and Douglas, 2002). It is a direct replacement of ‘one variable-at-a-time’ approach of experimentation, where experimenters vary only one variable at a time, keeping all other variables in the experiment fixed. In order to optimize the process parameters of both the first bond and second bond, a 20-run central composite design was used in this study. Based on the aforementioned experiment results, three levels of the factors including ultrasonic power, bonding force, and bonding time were determined. Table 1 shows the parameter levels for the central composite design. In this paper, pull force of the wedge bonds was used as a response or quality characteristic, that is also the output we want to mea-
Table 2 – Details of the ultrasonic power, bonding time, bonding force, bonding responses (pull force), and standard deviation (StdEv) for each run of DOE for the first bonds Run no. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Power (mW) 325 260 325 325 260 195 195 195 325 195 260 260 260 150 260 260 260 370 260 260
Time (ms) 20 30 40 40 30 40 20 20 20 40 30 30 30 30 14 30 46 30 30 30
Force (gf)
Pull force (gf)
StdEv
35 40 35 45 40 35 45 35 45 45 40 40 40 40 40 32 40 40 48 40
22.8 22.3 22.0 19.4 20.6 20.9 18.6 18.9 19.0 20.6 20.7 22.2 21.6 20.9 19.1 21.1 19.0 19.6 19.6 21.1
2.104 1.905 2.36 1.172 1.993 1.743 2.27 2.375 2.14 1.954 0.948 0.964 1.199 1.973 2.181 2.510 1.443 1.468 1.007 2.876
182
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 179–186
Table 3 – Details of the ultrasonic power, bonding time, bonding force, bonding responses (pull force), and standard deviation (StdEv) for each run of DOE for the second bonds Run no. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Power (mW) 325 260 325 325 260 195 195 195 325 195 260 260 260 150 260 260 260 370 260 260
Time (ms)
Force (gf)
20 30 40 40 30 40 20 20 20 40 30 30 30 30 14 30 46 30 30 30
35 40 35 45 40 35 45 35 45 45 40 40 40 40 40 32 40 40 48 40
Pull force (gf) 16.1 18.0 15.4 20.2 19.8 10.0 10.1 5.9 20.4 9.1 19.2 19.5 16.4 7.4 9.8 13.2 17.2 20.5 17.9 19.8
StdEv 3.75 3.44 4.46 3.02 2.14 3.05 3.495 2.424 1.62 3.17 3.41 3.66 2.47 3.95 3.55 6.03 5.71 2.86 2.77 1.32
sure during the experiment. Tables 2 and 3 show details of the ultrasonic power, bonding time, bonding force, bonding responses (pull force), and standard deviation (StdEv) for each run for both the first and second bonds. It can be found that for the copper wire used in this study, the pull forces of 20 gf or greater can be obtained for both the first and second bonds. The higher pull force and lower standard deviation of the first bonds can be achieved compared with the second bonds. Fig. 5 shows the contour plots of the pull force when the bonding time is 30 ms. It can be found that the highest pull force of the first bond was achieved with high power and low force. However, for the second bond, for the highest pull force, both high power and high force were required. This might be because of the tail formation with the wire clamp which closes and pulls the wire to break it at the heel of the second bond that requires more force and power. However, there is no pulling force of the clamp on the first bond, as shown in Fig. 1. According to the contour plots, the optimized ultrasonic power, bonding force, and bonding time for both the first bond and the second bond could be achieved, which are 260 mW/35 gf/30 ms and 325 mW/40 gf/30 ms, respectively. Fig. 6(a) shows the wedge bonds obtained with the optimized processing parameters when the bonding time is 30 ms with first bonds at a power of 260 mW and bonding force of 35 gf, and the second bonds at a power of 325 mW and bonding force of 40 gf. Fig. 6(b) shows the first bonds obtained with power 370 mW, force 40 gf, and time 30 ms. It could be found that with the increase of ultrasonic power, the deformation of the wedge bond increased, and excessive deformation of the first bonds occurred when a higher ultrasonic power was applied.
Fig. 5 – Contour plots of pull force vs. bonding power and bonding force when bonding time is 30 ms. (a) First bond and (b) second bond.
3.3.
Microstructure and hardness of the wedge bond
The possibility of using Cu wires bonded to Au/Ni plated Cu substrate has led to interest in the reliability of this metallurgical system. Fig. 7 shows the cross-sections of second bonds when the bonding force is 35 gf, bonding time is 30 ms and ultrasonic power is 260 and 370 mW, respectively. With the bonding power increasing, the thickness of the gold layer at the center of the Au/Ni metallization decreased, as shown in Fig. 7b, this will be discussed in detail at the following section. It seems that the ultrasonic power contributed to increased deformation of the copper wire because higher ultrasonic power made the wire softer due to ultrasonic softening. Fig. 8 shows two lines scan from the same cross-sections of Fig. 7a and b, which indicates that there was no obvious interdif-
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 179–186
183
Fig. 8 – Results of line scan from the same cross-sections of Fig. 5: (a) line scan from line 1 of Fig. 5a and (b) line scan from line 2 of Fig. 5b. Fig. 6 – Micrographs of the wedge bonds when the bonding time is 30 ms (a) the wedge bonds obtained with the optimized processing parameters, the first bonds (260 mW, 35 gf); the second bonds (325 mW, 40 gf); (b) the first bonds with high ultrasonic power shows excessive deformation (370 mW, 35 gf).
fusion between Cu and Au at these two bonding interfaces, however, a little amount of Au elements inside the Cu wire was found, which might be caused by the wear action and mechanical mixing.
Fig. 9 shows fractured surfaces after shear testing. Both of the fracture surfaces showed dimples which indicated desirable ductile bonded joints. The composition of the fracture surface was Au and Cu, which showed that the fracture occurred along the bonded interface between the Cu and Au layer. For 325 mW bonds, top layer of Au/Ni metallization lifted off somewhere and the Cu substrate was exposed during shearing, which could be found from EDX result of point A. Point B and point C show good bonding and mixing of the ele-
Fig. 7 – Cross-sections of second bonds with different bonding power showing thinning of the Au layer occurred with ultrasonic power increasing. (a) 260 mW, 35 gf, 30 ms and (b) 370 mW, 35 gf, 30 ms.
184
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 179–186
Fig. 9 – Fractured surface of the wedge bond after shear test (a) and (b) 325 mw; (c) and (d) 370 mw.
ments between wire and metallization, as shown in Fig. 9b. For the bonds with increased power of 370 mW, a small white particle was found at the fractured surface, as shown in Fig. 9c and d. The EDX result of point D shows that the particle consists of Cu and Au. There are three intermetallic phases (Cu3 Au, CuAu, CuAu3 ) in the binary Cu–Au phase diagram when the temperature is beyond 200 ◦ C. According to the literature, Cu moves rapidly through the Au film by boundary diffusion at temperatures of 100–300 ◦ C within approximately 1 h, and the diffusion coefficient for Cu in Au is D = 1.64 × 10−20 cm2 /s at 200 ◦ C (Hall and Morabito, 1978). This diffusion is faster than that of Au in Cu because the Cu atom is smaller than the Au atom. It was found from some ultrasonic wire bonding investigations that temperature rise at the bond interface was between 80 and 300 ◦ C (Ho, 2004). According to the Fick’s diffusion law, the thermal interdiffusion distance can be obtained from following equation: X2 = Dt, where t is the interdiffusion time, which is assumed as the bonding time 50 ms here. As a result of this, the thermal diffusion distance is not more than 0.3 A◦ . It was proposed that the interdiffusion between Al wire and Ni metallization was enhanced by ultrasonic vibration, and atomic diffusion between Au bond and Al metallization at a high level ultrasonic frequency (1.5 MHz) and bonding temperature 200 ◦ C was also found (Li et al., 2006). In this paper, the wedge bonding of the Cu wire was achieved at ambient temperature, and the interdiffusion between Cu and Au was not found from Fig. 8. Therefore it is concluded that the Cu diffusion into the Au during the wedge bonding at ambient temperature was negligible, and the formation of IMCs is not expected.
The above results confirmed that the mixing of the Cu and Au at the interface region and the achievement of the wedge bonding was caused by wearing action. In ultrasonic wedge bonding, the amplitude of the bonding tip oscillation is proportional to the applied ultrasonic power and the relative motion experienced when the wire is sliding, will lead to wear of material (or contaminant) according to an equation developed for contacting surfaces in relative motion (Peterson and Winer, 1980):
t=d
H 1 K PV
(1)
where t is the time required, d is the depth of material worn, P is the mean or nominal pressure, H is the hardness of the material, K is the wear coefficient constant, and V is the sliding velocity. This wear of material is termed fretting when small amplitude oscillations are involved. A sufficient removal of the contaminant layer is required for bonding to occur between the underlying metal surfaces. Bonding at room temperature is related to a wear mechanism induced by the ultrasonic vibrations. During bonding, the ultrasonic stick–slip friction causes friction power to be delivered at the interface. This power is partly transformed to mechanical wear. The wearing action breaks up contaminant and oxide layers allowing for areas of fresh metal of the opposing bonding partners to contact and bond to each other. From Fig. 9(a) and (c), the sheared fracture surface (footprint) showed the typical ellipse shape of wedge bonding. It is the wear caused by the relative motion that will allow intimate metal–metal contact and promotes subsequent bonding.
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 179–186
185
Fig. 10 – Microstructures of bonded joints when power is 325 mW, bonding force 35 gf, and time 50 ms (a) copper wire showing elongated grains in the middle (b) magnified picture showing the small equiaxed grains.
Fig. 11 – Microhardness test on the wedge bond when power is 260 mW, time 50 ms, and force 35 gf. (a) The first bond and (b) the second bond.
These regimes of relative motion combined with the stress distribution at the bonding interface caused by the bonding tool that account for the bonded footprint morphology. The rate and uniformity of the wear at the interface depends on the stress field amplitudes and uniformity at the interface, respectively. Because the stress is lowest right at the periphery of the contact, and is highest at the bonding interface directly below the tool. The thinning of the Au layer observed in Fig. 7b can be explained by the larger interfacial peak stresses right below the wedge tool where the thinning is observed when the higher ultrasonic power is used. Fig. 10 gives the microstructure of the bonded joint where ultrasonic power, bonding force, and time are 325 mW, 35 gf and 50 ms, respectively. The microstructures of the copper wire in the left upper part of Fig. 10a are elongated grains, which were produced during the drawing process of the wires when manufacturing. After wedge bonding, small equiaxed grains were found at the bonding interface region, as shown in Fig. 10b. As we discussed above, the temperature rising during room temperature wedge bonding process was not high enough to cause any microstructure change and even recovery of copper wire, thus it can be concluded that the recovery or recrystallization found in this study was due to the ultrasonic vibration. The horizontal ultrasonic vibration and the normal force of the wedge induced compressive stress and shear stress at the interfacial region, which resulted in the slipping and refining the grains the wedge bond. The slipping of the copper wire and refining grains provided a foundation of the further plastic deformation. The results were consistent with
the others studies in which the dynamic recovery or recrystallization of the aluminum wire was found during ultrasonic wedge bonding (Geissler et al., 2006; Krzanowski, 1990). Micrographs of microhardness test results are given in Fig. 11. The test was performed along the centerline of the copper wire. Five points for the wire part and the bonded joint part were tested. It could be found that the hardness at the deformed bonded joint part is a little higher than the wire part, which was probably caused by the strain hardening of the wedge bond. It was believed that the ultrasonic energy enhances the plastic deformation of the bonding wire due to the fact that the dislocations inside the wire absorbs the acoustic energy selectively, and the dislocations were activated from their anchoring locations, which makes the deformation of the bonding wire easier. After the acoustic energy was removed, the new defects made the bonding wire harder.
4.
Conclusions
(1) Copper wire bonding on an Au/Ni plated Cu substrate at ambient temperature was achieved, and the bonding parameters for both first bond and second bonds were optimized by design of experiment. To get strong pull force of the wedge bond, higher power and force were required for the second bond than for the first bond. (2) Cross-section analysis showed a continuous connection between the Cu wire and Au metallization when the
186
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 8 ( 2 0 0 8 ) 179–186
appropriate bonding parameters were chosen. With the increase of the ultrasonic power, the thickness of the gold layer at the center of the metallization was found. This is because the interfacial peak stresses occurred right below the wedge tool where the thinning is observed. (3) The interdiffusion between Cu wire and Au layer at the bonding interface during the wedge bonding at ambient temperature was assumed negligible. The achievement of copper wedge bonding was proposed to be the wear action and mechanical mixing induced by the ultrasonic vibration, and the ultrasonic power contributes to increasing deformation of the copper wire due to ultrasonic softening which was then followed by the strain hardening of the copper wedge bond, and the dynamic recovery or recrystallization of the copper wire caused by ultrasonic vibration during wedge bonding was also found. (4) Since the Au metallization layer has a good bondability, room temperature wedge bonding of the copper wire is feasible. However, if the bare Cu or Al substrates are used, it becomes very diffucult. In the future work, different kinds of the metallization layers should be investigated, and oxidation protection, coated copper wire, as well as the copper wire with different elongation should also be studied further.
Acknowledgments The research work was supported by National Natural Science Foundation of China under grant No. 50705021/E052104, MK Electron Ltd. of South Korea and the Development Program for Outstanding Young Teachers in HIT (Project No.: HIT. 2006: 01504489).
references
Geissler, U., Martin, S.-R., Lang, K.-D., Reichl, H., 2006. Investigation of microstructural processes during ultrasonic wedge/wedge bonding of AlSi1 wires. J. Electron. Mater. 35, 173–180.
Hall, P.M., Morabito, J.M., 1978. Diffusion problems in microelectronics packages. Thin Solid Films 53, 175–182. Harman, G.G., 1997. Wire Bonding in Microelectronics—Materials, Processes, Reliability, and Yield, 2nd edition. McGraw Hill, New York, NY. Harman, G.G., Albers, J., 1977. Ultrasonic welding mechanism as applied to aluminum-and gold-wire bonding in microelectronics. IEEE Trans. Parts, Hybr., Pack. 13, 406–411. Ho, J.R., 2004. Thin film thermal sensor for real time measurement of contact temperature during ultrasonic wire bonding process. Sens. Actuator A: Phys. 111, 188–195. Ho, H.M., Lam, W., Stoukatch, S., Ratchev, P., Vath III, C.J., Beyne, E., 2003. Direct gold and copper wires bonding on copper. Microelectron. Reliab. 43, 913–923. Krzanowski, J.E., 1990. Transmission electron microscopy study of ultrasonic wire bonding. IEEE Trans. Compon., Hybr., Manuf. Technol. 13, 361–368. Krzanowski, J.E., Razon, E., Hmiel, A.F., 1990. The effect of thin film structure and properties on gold ball bonding. J. Electron. Mater. 19, 919–924. Langenecker, B., 1966. Effects of ultrasound on deformation characteristics of metals. IEEE Trans. Son. Ultrason. 13, 1–8. Li, J.H., Wang, F., Han, L., Duan, J., Zhong, J., 2006. Atomic diffusion properties in wire bonding. Trans. Nonferrous Met. Soc. China 16, 463–466. Lum, I., Jung, J.P., Zhou, Y., 2005. Bonding mechanism in ultrasonic gold ball bonds on copper substrate. Metall. Mater. Trans. 36A, 1279–1286. Lum, I., Mayer, M., Zhou, Y., 2006. Footprint study of ultrasonic wedge-bonding with aluminum wire on copper substrate. J. Electron. Mater. 35, 422–433. Murali, S., Srikanth, N., Vath III, C.J., 2003. Grains, deformation substructures, and slip bands observed in thermosonic copper ball bonding. Mater. Charact. 50, 39–50. Peterson, M.B., Winer, W.O., 1980. Wear Theory and Mechanism from Wear Control Handbook. ASME United Engineering Center. Raymond, M., Douglas, C., 2002. Response Surface Methodology: Process and Product Optimization Using Designed Experiments, 2nd edition. John Wiley & Sons, Inc., New York. Takahashi, Y., Shibamoto, S., Inoue, K., 1996. Numerical analysis of the interfacial contact. Process in wire thermocompression bonding. IEEE Trans. Compon. Hybr. Manuf. Technol. Part A 19, 351–356.