Interfacial Reactions of Ni-doped SAC105 and SAC405 Solders on Ni-Au Finish during Multiple Reflows

Transcription

1 Interfacial Reactions of Ni-doped SAC105 and Solders on Ni-Au Finish during Multiple Reflows Toh C.H. 1, Liu Hao 1, Tu C.T 2., Chen T.D. 2, and Jessica Yeo 1 1 United Test and Assembly Center Ltd, 5 Serangoon North Ave 5, SINGAPORE Accurus Scientific Co., Ltd, , Wen Sien Road, Section 1 Jen-Der, Tainan, TAIWAN ch_toh@sg.utacgroup.com Abstract Solder-joint performances of and SAC105 with 200ppm and 500ppm Ni addition were investigated for electrolytic Ni-Au BGA pad finish. For each alloy system, ball shear tests, cross-sectional analysis and 3-D etching were performed to study the interfacial reactions after repeated reflows. Also, the effect of solder-mask pad design on solder joint integrity was investigated. In this study, all three systems gave rise to only (Cu,Ni) 6 5 IMC layer at the soldering interface after multiple reflows at 245 o C. The (Cu,Ni) 6 5 IMC had a needle-shaped structure after 1 time of reflow and the morphology remained the same after multiple reflows. appeared to be inappropriate for a BGA pad finish. It exhibited a drastic increase in the IMC layer thickness after 11 times of reflow and this corresponded to a significant increase in IMC fracture mode percentage. NSMD pad design led to ball pad lifting because the bonding between the pad and substrate was weaker than the bulk solder strength and the IMC-pad interface bonding. 1. Introduction Solder composition and ball-grid array (BGA) pad finish are some of the many factors that can affect intermetallic compound (IMC) formation at the interface. The IMC growth controls the strength of the solder joint. Excessive brittle intermetallics and weak interfaces can result in solder joint reliability concern leading to the BGA package failure. The adoption of SAC solders doped with nickel has steadily increased in the recent years, particularly for OSP finish. This is developed to replace the conventional lead-free alloys such as for drop reliability improvement by means of retarding the growth of Cu 3 IMC [1, 2]. In high volume manufacturing lines, it is always desirable to standardize solder balls composition for BGA packages with different pad finishes. However, there is limited report on the use of SAC solders doped with nickel for the Ni-Au finish, a popular industrial choice for pad finish. In the manufacturing assembly processes, multiple reflows are often required. During these processes, the solder joints are subjected to the repeated melting and solidification. Consequently, the microstructure of the BGA solder joints can be affected by the interfacial reactions such as dissolution of surface finish layers, compositional changes and IMC growth. The purpose of this study is to understand how and two nickel doped SAC solders react with the Ni- Au finish during multiple reflows. 2. Materials and Methods Test vehicles used in this study were BGA packages with Au/electrolytic Ni/Cu pad. The Au thickness is around 1um, while the Ni thickness is around 5um. Typical cross-section SEM image for the as-plated substrate are shown in Figure. 1. Electrolytic Au Fig. 1. Typical thickness for electrolytic Au plating Electrolytic Ni Cu Pad Solders balls with 0.4mm diameter from Accurus Scientific with three types of Pb-free composition were studied as given in Table 1. Both solder masks defined (SMD) and non-solder mask defined (NSMD) were studied. SMD pad refers to a ball pad defined by a solder-mask in which the copper pad is much larger. An NSMD pad has a solder mask opening that is larger than the copper pad area where the copper pad defined the solder ball structure. ICEPT2007 Proceedings 410

2 Table 1: Composition of Solder Balls Alloy Typical Composition (wt.%) 4% Ag, 0.5%Cu, balance 1% Ag, 0.5%Cu, 200ppm Ni balance 1% Ag, 0.5%Cu, 500ppm Ni balance Bulks solder area (ductile failure mode) (a) (b) IMC area (brittle failure mode) Solder balls were attached on the BGA substrates using a commercial flux. The layout of the package is shown in Figure 2. The reflow was accomplished using a hot air furnace equipped with 6 heating zones. Reflow temperature profile with 245 o C peak temperature and a time above the liquidus of approximately 46sec is shown in Figure 3. After 1 time of reflow, the packages were kept at a production floor at room temperature for around one month before underwent additional 3 times and 10 times of reflow at 245 o C. (c) (d) Fig. 4. Failure mode definition (a) bulk solders fracture with zero IMC fracture; (b) IMC fracture surface > 1% and <50%; (c)imc fracture surface > 50%; (d) pad lifting Specimens were cold mounted and cross-sectioned through a row of solder balls. The specimens were then ground with 2000-grit SiC paper, and mechanical polished using 0.3 & 0.05µm Al 2 O 3 powder. Micro-hardness tests were performed using 50gf load on the cross-section in the middle of a solder ball to examine the micro-hardness value after 1 time, and 11 times of reflows. Fig. 2. Substrate layout: FBGA 4x4mm-16B, 0.8mm pitch with 0.4mm solder ball diameter Temperature ( C) peak temperature=245 A HITACHI S-3000N scanning electron microscope (SEM) operating at 15KeV was used to study the interfaces and IMC microstructure. The SEM is equipped with an energy dispersive spectrometer (EDS) to analyze the IMC and phase composition. IMC thickness (in um) is defined as the ratio of IMC layer area to IMC layer length as illustrates in Figure 5. Area of IMC layer = A (µm 2 ) dw ell time above 217 =46 sec. Fig. 3. Reflow Time temperature (sec.) profile Fig. 3. Reflow temperature profile IMC layer Length of IMC layer = L (µm) Fig. 5. Definition for IMC layer s thickness A Dage 4000 tester was used for the high-speed ball zone shear test. A constant shear speed of 50mm/sec was applied. The gap between the substrate surfaces to the shear tool was kept at 30um. Fracture mode distribution was studied using optimal microscopy. Four levels of failure modes were defined as schematically shown in Figure 4. For each alloy composition & reflow conditions, about 80 solders balls from various BGA substrates were sheared. 3. Results and Discussion 3.1 Zone Balls Shears and Failure Modes Figure 6 plots the total number of IMC fracture for the various solders before and after multiple reflows. Only fracture mode for the first balls are reported and compared to avoid the neighboring balls sheared effect as schematically shown in Figure 7. Dark shaded bar represents more than 50% of the fracture surface (individual ball) exhibiting IMC interface fracture, whereas light shaded bar corresponds to less than 50% of the fracture surface is IMC fracture. ICEPT2007 Proceedings 411

3 Accumulation of IMC (%) failures 100% 80% 60% 40% 20% 0% 0%<IMC<50% 50%<IMC<100% Fig. 6. Accumulative IMC fracture percentage comparison after 1 time and 11times of reflow at 245 o C. with NSMD pad designs showed 100% pad lifting after 1 time of reflow. 3.2 IMC and Interfacial Microstructures The average thickness of IMC layer that formed at the interface of a solder ball and Ni-Au overplated Cu pad with reflow time is given in Figure 8. The IMC layer in, and systems was about 0.9, 0.7 and 1.2um, respectively after 1 time of reflow. exhibited a significant increase in IMC layer thickness up to 3um after 11 times of reflow while an average of 1.2um was maintained for and solders system. This corresponded to a significant increase in IMC fracture mode percentage as shown in Figure 6. The thick and brittle IMC layer may provide a low energy path for crack propagation. Shear tool Solder bump 1 st 2 nd 3 rd IMC layer thickness (µm) Average(µm) Standard deviation(µm) Pad 0.00 as-reflowed 4x reflows 11x reflows as-reflowed 4x reflows 11x reflows as-reflowed 4x reflows 11x reflows Fig. 8. IMC layer thickness comparison at as-reflowed, 4 times and 11times of reflow. Figure 9 shows that the bulk solder hardness for each solder system remained after multiple reflows. Indentations were carried in the middle of the solder ball, at least 50um away from the IMC layer. Fig 7. Zone balls shear failure mechanism. As the tool piece touches the first solder ball, the solder undergoes plastic deformation followed by fracture. The neighboring ball is then sheared by the first ball. The process is repeated until the whole row of solder balls is sheared. Microhardness (Hv50) Max Microhardness (Hv50) Min Microhardness (Hv50) Average Microhardness (Hv50) Stdev(Hv50) 0 Three failure modes were observed in this study, namely IMC fracture, bulk solder fracture and pad lift. The IMC fracture was with a shinny fracture surface, a typical character of brittle fracture. Fracture in the bulk solder was with a dull fracture surface depicting a ductile fracture. solders showed the least total number of IMC fracture after 1 time and 11 times of reflow at 245 o C. Although solders exhibited less total number of IMC fracture than after 1 time of reflow, the total number of IMC fracture for increased dramatically from 15% to 65% after 11 times of reflows at 245 o C. Fig. 9. Bulk solder micro-hardness comparison after 1 time and 11 times of reflows at 245 o C Typical cross-sectional SEM images for the, and system after 1, 4 and 11 times of reflows are shown in Figure 10 and Figure 11. EDS analysis indicated the IMC layer compositions at the soldering interface are consistent with the stoichiometry of the compound (Cu x,ni y ) 6 5. No other IMC was detected after multiple reflows at 245 o C. The (Cu x,ni y ) 6 5 is a uniform layer conforming to the nickel surface finish. ICEPT2007 Proceedings 412

4 significant growth in (Cux,Niy)65 IMC layer thickness after 11 times of reflow showed no morphology changed. The thick IMC shown in Figure 11 was attributed to the IMC layer thickening of (Cux,Niy)65. acicular (a) (as reflowed) acicular AuxCuyNiz1-x-y-z (b) (4x reflows) Ag3 Fig. 10. Solder joint microstructures comparison after 1 time of reflow@245oc (left) and 4 times of reflows@245oc (right) (c) acicular (d) (as reflowed) acicular Residual solder Au4 (e) (4x reflows) Fig. 11. Solder joint microstructures comparison after 11 times of Figure 12 shows the 3-D IMC images for each solder system. Deep etching revealed that the morphology of (Cux,Niy)65 were needled shaped compound protruding into the solder, making the interface very rough as seen in Figures Needle-shaped IMC morphology was reported by Lee et al. [3] for SAC305 on electroplated Cu/Ni/Au after 20 times of reflows. The morphology remained the same after repeated reflows at 245oC up to 11 times. It is interesting to note that a ring pattern of IMC with a relatively larger IMC was observed for and after 1 time and multiple reflows. solders which exhibited a ICEPT2007 Proceedings (f) Fig D-IMC images comparison for as-reflowed, 4 times and 11 times of reflow at 245oC. 413

5 (Cu,Ni)65 [4]. In the Cu-Ni- ternary system as shown in Figure 14, (Cu,Ni)65 is more stable than Ni34, (Cu,Ni)65 preferentially formed at the interface with the Cu in the solder [5]. In this study, only Cu65 with some solution of Ni exited at the interface up to 11 times of reflow for all alloys system. (g) (as reflowed) (h) (4x reflows) (i) Fig. 12 (con t). 3D-IMC images comparison for as-reflowed, 4 times and 11 times of reflow at 245oC. Fig. 14. Isothermal Cu--Ni phase diagram at 235oC. Redrawn from [5] As more reflow cycles were applied, the IMC were precipitated out in the solder matrix and thereby the amounts of Cu and Ni were reduced. Interestingly, the available of the Cu inside the solder matrix next to the IMC is seems to sustain the substantial growth of (Cu,Ni)65 after multiples reflows for. However, this is not the case for the identical but with 300ppm less Ni concentration. Further investigation will be required to understand the mechanism behind the much higher IMC growth for than after 11 times of reflows. Figure 10 and 11 showed the inside the bulk solders were pebble or plate-like. Figure 13 shows a typical 3-D image of for after 4 times of reflow. The had a prismatic shape and the sizes were in the ranges of 6-12um. Au4 platelet was detected in the bulk solders for after 11 times of reflow (see Figure 11). Also, a compound of AuxCuyNiz-x-y-z was found inside the bulk solder for after 4 times of reflow (see Figure 10). Residual solder 3.3 Effect of solder mask design In this study, solder balls with the SMD pad design exhibited either bulk solder fracture or IMC fracture while all those with NSMD pads design showed pad lift phenomenon. Figure 15 provides a side-by-side comparison of such pad designs at the same length scale. For SMD pad design, the bonding between solder balls and copper pads was provided by the top pad area. The solder resist overlapped the pad area and enhanced the adhesion between the copper pad and the substance. This helped to enhance adhesion of the copper pad, resulting in a ball shear failure mode. acicular Fig. 13. morphology for with Ni-Au surface finish after 4 times of reflow. During soldering, the Ni from BGA surface finish diffused into the solder and an IMC formed. If the solder did not contain Cu, the IMC at the interface was typically Ni34, however, when Cu existed as little as 0.6wt.% in the solder, the IMC became ICEPT2007 Proceedings 414

6 Fig. 15. Comparison of SMD (top) and NSMD pads (bottom) after balls soldering. In this study, the overplated Cu pad diameter is 0.33mm. Figure 16 shows a typical force-displacement diagram for a sheared single ball with SMD and NSMD pad design. The slope to failure was produced after removal of the neighboring balls. In this example with a 0.33mm pad diameter and a 0.4mm ball, SMD pad design exhibited slightly higher shear force than NSMD pad design. Also, the large area under the force-displacement diagram for SMD design indicates its ability to absorb more energy up to a fracture. A study by Lim et al.[6] has shown that the relative strength of SMD and NSMD ball pads was determined by the pad size and substrate thickness. It is also interesting to note that the diagram shows multiple stress peaks before failure for NSMD pad design. Force (g) NSMD_1 NSMD_2 NSMD_3 SMD_1 SMD_2 SMD_ Displacement (µm) Fig. 16. Typical force-displacement graph for single shear test. Comparison of SMD and NSMD pad design for solder. A constant shear speed of 500um/sec was applied. All three systems gave rise to only (Cu,Ni) 6 5 IMC layer at the soldering interface after multiple reflows at 245 o C. The (Cu,Ni) 6 5 IMC had a needle-shaped structure after 1 time reflow and the morphology remained the same after multiple reflows. appeared to be inappropriate for a BGA pad finish. It exhibited a drastic increase in the IMC layer thickness after 11 times of reflow and this corresponded to a significant increase in IMC fracture mode percentage. NSMD pad design led to ball pad lifting because the bonding between the pad and substrate was weaker than the bulk solder strength and the IMC-pad interface bonding. 5. Acknowledgement The authors would like to thank the management teams of United Test and Assembly Test Center Ltd (UTAC) and Accurus Scientific Co., Ltd for their support on this project. 6. References 1. J.Y. Tsai, Y.C. Hu, C.M. Tsai and C.R. Kao, A study on the reaction between Cu and 3.5Ag solder doped with small amount of Ni, J. Electronic Materials, Vo1. 32, No.11, 2003, pp I.E. Anderson, J.C. Foley, B.A. Cook, J. Harringa, R.I. Terpstra and O. Unal, Alloy effects in near-eutectic -Ag- Au solder alloys for improved microstructral stability, J. Electronic Materials, Vol.30, No.9, 2001, pp KY. Lee, M Li, D.R. Olsen and W T. Chen, Microstructure, Joint Strength and Failure Mechanism of - Ag, -Ag-Cu versus -Pb-Ag Solders in BGA Packages, 2001 ECTC. 4. C.E Ho, R.Y. Tsai, Y.L. Lin and C.R. Kao, Effect of Cu Concentration on the Reactions between -Ag-Cu Solders and Ni, J. Electronic Materials, Vol. 31, No , pp K. Zeng and K. N. Tu, "Six cases of reliability study of Pbfree solder joints in electron packaging technology," Materials Science and Engineering Reports, R38, (2002). (A review paper). 6. A.C.P. Lim, T.K Lee, and Airin Alamsjah, The Effect of Ball Pad designs and substrate materials on the performance of second-level interconnects, 2003 EPTC, pp Conclusions The following conclusions could be drawn for the, and solders when reacted with Ni-Au surface finish. ICEPT2007 Proceedings 415