SEM micrograph showing the fracture plane of the solder joints after 15 cycles of reflow of (a) Sn – Zn – 3Bi and (b) Sn – Zn – 1Bi. 

Context in source publication

Context 1

... 12 The ball grid array (BGA) package has achieved wide- spread application by virtue of enhancing miniaturization and performance. Along with the progressive develop- ment of electronic interconnection toward smaller size and higher integration, the mechanical integrity of solder joints is a serious reliability concern. So far, substantial studies of the mechanical strength of BGA solder joints have been made with respect to different solders, pad and substrate designs, shearing/pulling parameters, reflow ambience, and aging conditions. However, few studies have concentrated on the effect of multiple reflow cycles, 13–17 despite the fact that BGA solder joints may undergo several cycles of reflow during component manufacturing, soldering, and rework processes. Thus, this study aims to investigate the effect of repeated reflow cycles on the reliability of Sn–8Zn–1Bi and Sn– 8Zn–3Bi (wt%) solder joints on Au/Ni metallization. A solder mask defined copper bond pad on the flexible substrate of a BGA package was used as a base for elec- trodeposition of Ni and Au. The solder mask-opening diameter was 0.6 mm with 7–9 ␮ m thick Ni at the ball pad. The thickness of Au layer was 0.5–0.9 ␮ m. The compositions of the solder alloys were Sn–8Zn–3Bi and Sn – 8Zn – 1Bi (wt%). The Sn – Zn – 3Bi solder was identified by differential scanning calorimetric (DSC) measurements to have a melting range of 187 – 197 ° C. The solidus and the liquidus temperatures of the Sn – 8%Zn – 1%Bi solder alloy were around 192 and 199 ° C, respectively. Lead-free solder balls with a diameter of 0.76 mm were placed on the prefluxed Au/Ni/Cu bond pad of the substrates and reflowed at a temperature of 230 ° C for 1 min in a convection reflow oven (BTU VIP-70N, BTU International Inc., North Billerica, MA). The flux used in this work was a commercial no-clean flux. The as- sembled packages experienced 2 – 25 reflow cycles to ex- amine the interfacial reactions between the solder and the thin-film UBM. Shear tests were performed on both the as-reflowed and multiply-reflowed samples using a Dage Series 4000 Bond Tester (Dage Holdings Ltd., Aylesbury, UK). The shear tool height and the test speed of the shear test in this work were about 40 ␮ m and 300 ␮ m/s, respectively. About 40 randomly chosen solder balls were sheared to obtain the average and the extent of deviation. The fracture surfaces after the ball shear tests were investigated thoroughly by scanning electron microscopy (SEM) in the secondary electron mode as well as by energy- dispersive x-ray (EDX). To investigate the microstructures, the as-reflowed and multiple-reflowed samples were mounted in resin, cured at room temperature, mechanically ground, and then polished to obtain the cross sections of the solder/UBM interfaces. The chemical and microstructural analyses of the gold-coated cross-sectioned samples were obtained using a Philips XL 40 FEG (Eindhoven, The Nether- lands) scanning electron microscope equipped with an EDX spectrometer. The accuracy of the compositional measurement was about ±5%. To determine the formula composition of the IMCs, the chemical analyses of the EDX spectra were corrected by standard atomic number, absorption, and fluorescence (ZAF) software (EDAX International, Mahwah, NJ). The back-scattered electron (BSE) imaging mode of the SEM was used for the interfacial study. The microhardness of the polished samples was measured in a LECO M 400H hardness- testing machine (LECO Corporation, St. Joseph, MI) with a load of 200 g and an indentation time of 15 s. The interfacial reactions between the Sn – Zn – Bi solders and electrolytically deposited Ni/Cu bond pads were conducted at 230 ° C. The mechanical strength of the interface was measured for each reflow condition. The change of the average shear load and their standard deviations are shown in Fig. 1. It is noted that the average shear strength of the Sn – Zn – 3Bi solder joints was higher than that of the Sn – Zn – 1Bi solder joints. The shear strength was found to increase with a larger number of multiple reflow cycles. After 10 reflow cycles, the strength did not change much for both the solder alloys. After measuring the shear loads, fractured surfaces of the residue pads and sheared balls were immediately studied by SEM. The results reported here (from fracture surface and cross-sectional studies) are based on the highest number of similar occurrences (i.e., mode, in the language of statistics) from each readout point of shear strength data. Generally, there were two main failure modes, I and II, observed after the shearing tests, as schematically shown in Fig. 2. Mode I was identified to be completely cut through the bulk solder balls as illus- trated in Fig. 2(b). This fracture occurred at a location near but lower than the shearing height (40 ␮ m), leaving a thick layer of solder on the pad. This indicates that the solder/pad bond is much stronger than the shear strength of the bulk solders. Figure 3 shows the fracture plane of the solder joints. Mode II is entirely different, in that the whole pad is lifted-off, as depicted in Fig. 2(c). When the number of reflow cycles was more than 20, some of the shearing failures were found to be mode II, regardless of solder type. It is interesting to note that the lift-off of the pads was noticed in a particular area. Due to repeated heating, the textures of the polymide materials might degrade. After reflow, the presence of voids/delamination was checked by scanning acoustic microscopy (SAM). In some area of the substrates, delamination of the polimide layer from the Cu heat sink was observed after 25 reflow cycles. Increasing power dissipation and decreasing die sizes drive the increasing need to attach heat sinks to packages. A wide variety of adhesive materials is used for this application. It is critical to maintain a uniform, thin, void-free bond line for the lowest thermal resistance. During repeated heating and cooling, the thermal stress generated due to coefficient of thermal expansion (CTE) mismatch will create delamination at the site of very small defects (e.g., voids). Therefore, the whole solder joint is maintained intact, while the resin under the pad is torn apart from the Cu heatsink. Figure 4 shows a delaminated area in a substrate after 25 reflow cycles. It was confirmed that for up to 20 reflow cycles, the dominant shearing behavior for both the Sn – Zn – Bi solder joints was mode I. The fracture surfaces of the two solder joints were more or less similar, showing a large ductile deformation of the solder. It is suggested that repeated reflow cycles enlarge the contact area and lower the height of the solder joints. As a result, the forces become higher because of the larger sheared area. A schematic diagram of the extent of deformation of the bulk solder is shown in Fig. 5. Due to the solder mask, the contact area between the solder and the bond pad remains the same. However, the area along the shearing height is increased after multiple reflow cycles [Fig. 5(b)]. It is also suggested that the increase in shear load of the solder joint might be related to the strength- ening effect of the solder alloy due to homogenization at the time of reflow. 18 Detailed cross-sectional studies were carried out to investigate the relationship between the shear strength and the interfacial morphologies of the Sn – Zn – Bi solders with the Au/electrolytic Ni/Cu pads. These are cross-sectional micrographs with the section plane per- pendicular to the interfaces. Both interfaces, more or less, reveal similar features: the solidified solder, reaction zone, original electrolytic Ni layer, and Cu pad. Interest- ingly, Au did not dissolve into the solders. Instead, Au formed an intermetallic compound at the interface. EDX analysis of the reaction zone revealed that the IMC is composed of Au and Zn. EDX results shows that the Au percentage of this layer is about 25 at.%. This observa- tion implies that the IMC layer is the AuZn 3 compound. In the electrolytic Ni/Sn – Zn – Bi solder joint, the thickness of intermetallics was around 3 – 5 ␮ m. As the initial IMC contained no Ni, the consumption of the original Ni layer in both the Sn – Zn – Bi solder joints was negligible. Just after the initial reflow cycle, layer-type spalling was observed in both the solder systems. At the interface, a very thin layer of IMC was noticed. Figure 6 shows the microstructures of the two solder interfaces after 5 times of reflow. Even after 5 reflow cycles, the interfacial IMC is still very thin. According to EDX, the IMC at the interface after 10 reflow cycles consists of Au, Ni, Sn, and Zn. The average composition of the interfacial IMCs layer near the substrate side was determined to be 52 – 59 Zn, 10 – 12 Au, 9 – 15 Ni, and 16 – 22 Sn (at.%). After 25 reflow cycles, the spalled Au-Zn IMC from the interface maintains its integrity as a layer in both of the solder systems (Fig. 7). The Au – Zn layer-type IMC in the bulk was situated near the interface even after such a large number of reflow cycles. The average thickness of the IMC at the interface was around 1.2 ␮ m, and thus the growth rate of the Au – Ni – Sn – Zn IMCs was much reduced. By measuring the remaining Ni thickness from SEM micrographs and by subtracting this from the initial thickness, the consumed Ni thickness was deduced. Only about 0.6 ␮ m of the Ni layer takes part in the reaction with both the Sn – Zn – Bi molten solders after 25 reflow cycles. The participation of the Ni with the interfacial IMC is the source of the consumption of the electrolytic Ni layer. Although the IMC spalling started very early in the reflow cycles, it did not greatly influence the shear strength of the solder joints. The highly reactive nature of the Zn confirms an instant IMC formation at the interface with the spalling of the Au – Zn compound layer. The interfacial IMC together with the unreacted Ni provides the adhesion ...