Micro- Test on Lead-Free BGA Balls on Au/Electrolytic Ni/Cu Bond Pad Shengquan Ou*, Yuhuan Xu and K. N. Tu Department of Materials Science and Engineering, UCLA, Los Angeles, CA, 90095-1595 M. O. Alam, and Y. C. Chan Department of Electrical Engineering, City University of Hong Kong, Kowloon, Hong Kong ousq@ucla.edu Abstract The most frequent failure of wireless, handheld, and movable consumer electronic products is an accidental drop to the ground. The impact may cause interfacial fracture of wire-bonds or solder joints between a Si chip and its packaging module. Existing metrologies, such as ball shear, and pull test cannot well represent the shock reliability of the package. In our study, a micro-impact machine is utilized to test the impact reliability of three kinds of lead-free solders: 99Sn1Ag, 9.5Sn1Ag0.5Cu and 97.5Sn1Ag0.5Cu1In (in weight percent, later abbreviated as Sn1Ag, Sn1Ag0.5Cu, and Sn1Ag0.5Cu1In in the paper). The effect of thermal aging on the impact toughness is also evaluated in this study. We find a ductile-to-brittle transition in SnAg(Cu) solder joints after thermal aging. The impact toughness is enhanced by the thermal aging. This is a combination effect of the growth of intermetallic compound (IMC) at the interface provided strong bonding, and the softening of the solder bulk during the thermal aging provided more plastic deformation. Introduction The drop impact induced solder joint fracture has become one of the critical system failure modes of interest in the electronics industry due to the migration of market focus to portable applications. For example, cellular phones might be broken due to impact by dropping. Especially, in recent years, while the weight of the personal digital assistant (PDA) devices is reduced greatly, ball grid arrays (BGAs) and chip size packages (CSPs) is adopted to effectively reduce mounting areas. The major damage of the BGA/CSP solder joints is not from external temperature changes but short-time stresses like drop impacts.[1] To evaluate the joint reliability between BGA ball and under bump metallization (UBM) pad, industry widely uses shear test and pull test.[2] Unfortunately, these tests can not evaluate the impact reliability of solder joints, because the testing speeds are typically lower than 1 mm/s, well below the velocity of impact applied to solder joints by dropping. Some previous study by Chiu et. al.[3] proved that there is a very strong correlation between drop reliability and voiding at the UBM/solder interface. However, ball shear testing does not correlate to drop test performance, and ball pull strength is not a good indicator of shock reliability either. Therefore, how to characterize the impact reliability induced by dropping becomes very crucial. Recently, the research group from Hitachi Metals, Ltd., Japan[,5] proposed a miniature impact test for solder bumps by adopting the principle of the classic Charpy impact test.[] Based on their work, we have built a micro-impact testing machine to study quantitatively the bonding strength between BGA ball and UBM pad (as shown in Fig. 1). The sample is placed on an XYZ-adjustable positioning stage. The initial position of the hammer and its final position after impact are recorded by an angle recorder with accuracy of ±0.5 degree. Since the measured angle difference in a typical impact test is larger than degree, the resolution is about 5% of the measured energy change. Sn1Ag, Sn1Ag0.5Cu and Sn1Ag0.5Cu1In are chosen for this study because they have good wettability and mechanical strength. Adding Indium into SnAgCu solder can lower the melting temperature by about C, which can reduce the temperature distribution across the board, especially large boards. The UBM structure on the bond pad is Au/electrolytic Ni/Cu metallization. tests are carried out after reflow and after the thermal aging at 150 C. Fig. 1 Photograph of the micro-impact test machine. Experimental A. Sample preparation We used 70 µm (in diameter) size solder balls for the impact test. The opening of the solder resist was 50 µm in diameter, 30 µm in depth at the Au/electrolytic Ni/Cu bond pad. Gold thickness over electrolytic Ni was 0.5 µm. The 0-703-90-9/05/$.00 05 IEEE 7
Fig. 2 JEDEC reflow profiles for Sn-Pb and Pb-free assemblies. average thickness of electrolytic Ni was 2.5 µm, and the thickness of Cu underneath the electrolytic Ni was 13 µm. The reflow process of the solder ball on bond pad followed the standard JEDEC reflow profiles for Pb-free assemblies as shown in Fig.2. The reactive mild active (RMA) flux was used during reflow, and the peak reflow temperature for Sn1Ag and Sn1Ag0.5Cu was 0 C, and it was 0 C for Sn1Ag0.5Cu1In. The duration time at the peak reflow temperature was 1 minute. After reflow, some substrates were put into furnace for thermal aging at 150 C. We took samples out at 0 hour, 500 hour and 00 hour to test the change of the impact toughness with the aging time. The fracture surface and crosssectional microstructures of the solder joints were analyzed by optical microscope, scanning electron microscopy (SEM) and energy dispersive X-rays (EDX). The fracture surface was examined on both the solder ball side and the bond-pad side for complete understanding of the failure mechanism. B. test on solder bumps The way to determine the impact toughness was shown in Fig. 3(a). The hammer hit one side of the solder ball, and created two fracture surfaces, and at the same time, deformed the solder bump. The energy spent was measured by the gravitational potential energy change of the hammer before and after the impact. The potential energy change, E, was measured from the difference between the initial pendulum height, h 0, and the maximum height achieved during the follow-through, h, or the difference between the height of the hammer before and after the impact. E = mgh 0 mgh = mg h where m is the mass of the hammer (in units of gram), g is gravity constant or acceleration of gravity (90 cm/sec 2 ), h = h 0 h = L ( cosθ 0 cosθ ), where L is the length of the hanging hammer (in units of cm), and θ 0 and θ are the angles of the hammer before and after impact. The unit of the potential energy is in mj, mille Joule. The size of the substrate we used for the impact test was 3.5 cm 3.5 cm. The positions of the solder balls were specially designed for the impact test. Fig. 3 showed the designed pattern. On one substrate, there were balls in total. Sample # Solder ball No solder Total solder ball number = Fig. 3 (a) a solder ball: 1-- start position of hammer, 2 impact to the ball, and 3 final position of the hammer. Designed pattern of solder balls arrays on the substrate. Results & Discussion A. Microstructures of solder bumps on Au/Ni/Cu pad As shown in Fig., we can see, after reflow, a very thin layer of binary Ni 3 Sn IMC formed at the Sn1Ag solder and Au/electrolytic Ni/Cu interface. By adding 0.5 wt.% Cu into Sn1Ag solder, the IMC at the interface changed into the ternary compound (Cu,Ni) Sn 5 due to the gain in free energy. [7] At the Sn1Ag0.5Cu1In solder and Au/Ni/Cu interface, we
150 C/00 As-reflowed was defined as a brittle fracture. In this study, we also found a third mode of fracture within the bond pad metallization. This mode would occur after a long time aging. (d) (a) Ni3Sn Ni3Sn (e) (f) (c) Cu-Ni-Sn Fig. 5 Schematic of the ductile-to-brittle transition behavior in BBC metals. (after ref. 9) Fig. SEM pictures of the cross-section of the solder joints. (a) and (d) Sn1Ag; and (e) Sn1Ag0.5Cu; (c) and (f) Sn1Ag0.5Cu1In. found the formation of very tiny needle-like Cu-Ni-Sn compound. The IMCs at the interface grew slowly during the thermal aging process. In the solder bulk, the coarsening of Ag3Sn compound has occurred. It was hard to tell in Fig. because of the chemical etching. But this was verified by the work done by M. Date. [] Due to the low Ag concentration (1 wt.%) in the solder composition, we did not observe large plate-like Ag3Sn after the thermal aging, which was known to be detrimental to the joint strength. There was also an expected grain growth after the long time aging at this high temperature for solder joints. B. toughness of the solder joints One of the primary functions of impact test is to determine whether or not a material experiences a ductile-to-brittle transition. It is known that in many body-centered-cubic metals, including ferretic steels, a ductile-to-brittle transition temperature (DBTT) exists. The metals are ductile above DBTT, yet they become brittle below DBTT, as shown in Fig. 5.[9] From the impact test on solder joints, we also found a ductile-to-brittle transition in the solder joint. This ductile-tobrittle transition found in our experiments, was not due to application temperature, rather it was due to reflow, or aging time, or a combination of the two, or the bond pad metallization. Using SEM and optical microscope, we examined both solder ball side and bond pad side to evaluate the mode of fracture. In Fig., we showed all the fracture modes observed in our impact tests. Mode 1 was the fracture across the solder bump. We treated this mode as ductile fracture. Mode 2 was the fracture across an interface, which Ductile Brittle Bond-Pad Fig. modes in the impact test on solder joints. A statistical analysis of the fracture modes of all solder joints under every condition was summarized and plotted in Fig. 7. From the plot, we can tell that all the as-reflowed Sn1Ag solder joints had ductile fracture across the solder bulk. With thermal aging, the fracture mode shifted to brittle fracture. A clear ductile-to-brittle transition phenomenon illustrated in this case. The transition also appeared in Sn1Ag0.5Cu solder joints, but less obvious. EDX analysis was employed to get the composition information on the collected solder balls, which had been knocked out from the bond-pad. 9
Percentage, % 0 0 0 0 Sn1Ag Sn1Ag0.5Cu Sn1Ag0.5Cu1In solder joints, the impact toughness increased graduately to the point of 500 hours aging at 150 C. After 500 hours aging, the impact toughness started to drop. In both cases of Sn1Ag0.5Cu and Sn1Ag0.5Cu1In, we can see an increase of the impact toughness with the aging time. 0 Reflow 0hrs 500hrs 00hrs Reflow 0hrs 500hrs 00hrs Reflow 0hrs 500hrs 00hrs Thermal Condition Thermal Condition Thermal Condition Fig. 7 Percentage of different fracture mode for Sn1Ag, Sn1Ag0.5Cu, and Sn1Ag Sn1Ag0.5Cu1In solder joints at each thermal condition. The results from EDX analysis (listed in Fig. ) indicated the fracture crossed interface was located in the place between Cu-Ni-Sn compound and the UBM metallization layer. During aging, the thickness of IMCs at the interface increased with time. The brittle nature of IMCs resulted in a fracture across the interface. Lots of studies [7, ] on the interfacial reaction between Ni and solders reported Kirkendall voids formation at aging due to unbalance of out-flux and in-flux. As a result, Kirkendall voids would cause a poor adhesion between the metallization layer and the solder joints. All these factors leaded to the brittle fracture at the interface. In the case of Sn1Ag0.5Cu1In solder joints, they appeared that all interface fracture thru all thermal condition. This may related to the tiny needle-like Cu-Ni-Sn compound formed at the interface, and higher yield stress compared with Sn1Ag and Sn1Ag0.5Cu. The needle-like IMC may provide more channels for fast Ni diffusion. But we need further detailed study on the cross-section to prove this. Bond-Pad Side Solder Ball Side Element Weight% Atomic% ------------------------------------------- P 0.0 1. Ni.3 1.21 Cu 1. 2.0 Ag 0. 0.75 Sn 70.90 5.79 Total 0.00 Element Weight% Atomic% ------------------------------------------- P 0.00 0.00 Ni 1.092.1 Cu 1.7 3.0 Ag 0.00 0.00 Sn 97.05 9. Total 0.00 Toughness, mj Toughness, mj Toughness, mj 1 1 1 Sn1Ag 0 0 00 00 00 00 Sn1Ag0.5Cu 0 0 00 00 00 00 Sn1Ag0.5Cu1In Fig. Typical fracture across the interface in Sn1Ag solder joints after aging at 150 C for 00 hours. The results of the change of the impact toughness with the aging time were shown in Fig. 9. We found as for Sn1Ag 70 0 0 00 00 00 00 Fig. 9 toughness change as a function of thermal aging time for Sn1Ag, Sn1Ag0.5Cu, and Sn1Ag0.5Cu1In solder joints.
In the impact test, if we assume that the impact energy is the sum of the energy spent to create the fracture surfaces and the energy spent to deform the solder ball, we have E total = E fracture surfaces + E deformation of solder ball We should emphasize that in the ductile fracture of the solder joint as shown in Fig. (a), the solder ball itself, which was knocked away, should have been heavily deformed plastically. Therefore, more impact energy might have been spent to deform the solder ball than to create the fracture surfaces. On the other hand, when brittle fracture occurred as shown in Fig., we expected that the solder ball itself might have less plastic deformation than that in ductile fracture shown in Fig. (a). Generally speaking, if the matrix of solder ball is soft, it absorbs more impact energy and deforms more heavily. On the other hand, if the matrix of the solder ball is rigid, it will not absorb much impact energy and will deform much less. Bond-Pad Side Solder Ball Side (a) (c) Fig. Three kinds of failure shapes in impact test: (a) fracture in solder bulk; fracture at interface; (c) fracture at interface and a deformed solder bump During the long time thermal aging, the IMC growth at the interface could strengthen the bonding before continuous Kirkendall voids formed. In addition, the grain growth in the solder matrix softened the solder bumps by reducing the total area of grain boundary. Therefore, we observed huge deformation in the solder bulk as shown in Fig. (c). Combine those factors, the energy needed to create the interfaces between IMC and solder and deform the solder bump would increase. In another word, the impact toughness would expect to increase for longer aging time. Conclusions In this study, we find impact test is an effective tool to detect the ductile-to-brittle transition in the Sn1Ag, and Sn1Ag0.5Cu solder joints on Au/electrolytic Ni/Cu bond-pad. The fracture interface was located at the place between UBM metallization layer and Ni(Cu)-Sn intermetallic compound. The needle-like Ni-Cu-Sn compound formed at the Sn1Ag0.5Cu1In and Au/electrolytic Ni/Cu interface might be the season for the weak interface bonding at as-reflowed condition. The change impact toughness with the thermal aging time showed an increasing trend. The continuous IMC formation at the interface without Kirkendall voids formation and the soften solder bumps after aging might contribute to the increase of impact toughness. References 1. Kinuko Mishiro, et al, Effect of the drop impact on BGA/CSP package reliability, Miroelectronics Reliability, 2 (02), pp. 77-2. 2. Nishiura M, Nakayama A, Sakatani S, Kohara Y, Uenishi K, Kobayashi KF., Mechanical strength and microstructure of BGA joints using lead-free solders, Mater Trans., 3 (02), pp.02. 3. Tz-Cheng Chiu, Kejun Zeng, Roger Stierman and Darvin Edwards, Kazuaki Ano, Effect of Thermal Aging on Board Level Drop Reliability for Pb-free BGA Packages, The proceedings of the 5th ECTC, Las Vegas, NV, 0, pp.5.. Shoji T, Yamamoto K, Kajiwara R, Morita T, Sato K, Date M., Proceedings of the th JIEP Annual Meeting, 02, pp.97. 5. M. Date, T. Shoji, M. Fujiyoshi, K. Sato, and K. N. Tu, Ductile-to-brittle transition in Sn-Zn solder loints measured by impact test, Scripta Materialia, 51 (0), pp.1.. John M. Holt, editor, Charpy Test: Factors and Variables ASTM STP 72, Philadelphia, PA, (1990). 7. K. Zeng and K. N. Tu, Six cases of reliability study of Pb-free solder joints in electronic packaging technology, Mater. Sci. Eng., R., 3 (02), pp. 55.. M. Date, T. Shoji, M. Fujiyoshi, K. Sato, and K. N. Tu, " Reliability of Solder Joints," The proceedings of the 5th ECTC, Las Vegas, NV, June 0, pp. -7. 9. J. P. Schaffer, A. Saxena, S. D. Antolovich, T. H. Sanders, Jr., and S. B. Warner, The Science and Design of Engineering Materials, 2rd Ed., WCB McGraw-Hill (1999), pp. 32.. M. O. Alam, Y. C. Chan, and K. N. Tu, "Effect of 0.5 wt. % Cu addition in the Sn-3.5%Ag solder on the interfacial reaction with Au/Ni metallizaion," Chemistry of Materials, 15 (03), pp.30-32. 71