Creep of Sn-3.5Ag Flip-Chip Solder Joints in Shear

Flip-Chip Sn-3.5Ag Shear Data

Shear data for Sn3.5Ag solder joints was obtained from two sources:

• Shear creep tests conducted by Wiese et al. (2001) on hour-glass shaped, flip-chip solder joints of silicon-on-silicon Sn-3.5Ag assemblies.

• Solder bumps were deposited using paste and processes from two manufacturers. Pads were 0.1 mm x 0.1 mm or 0.2 mm x 0.2 mm and the hour-glass shaped solder joints were 0.15 mm or 0.2 mm in height. The average shear stress was defined as the applied force divided by the narrow area or minimum cross-section area of the hour-glass shaped solder joints.
• Test results given in Figure 19 in Wiese et al. (2001) were converted from shear to tensile data. The digitized raw data is given in Table A.7 in Appendix A. The creep tests were run at 5°C, 10°C and 50°C. Bump paste was from two manufacturers identified by the labels DG and PT. In Table A.7, we re-converted the tensile data to shear using the same Von Mises transformation as was used by Wiese et al. (equations (9) and (10) in Wiese's paper).

• Shear creep tests by Yang, H. et al. (1996):

• Flip-chip joints from silicon-on-silicon assemblies (33 x 33 I/O) using standard reflow were apparently barrel-shaped with a minimum load bearing area at the joint to chip interface.
• Limited isothermal test data was available at 25°C and 80°C. The digitized raw data, from Figure 12 in Yang, H. et al (1997), is listed in Table A.8.

Comparison of Flip-Chip and CCC Solder Joint Shear Data

Figure 20: Plot of Sn3.5Ag flip-chip and CCC joint shear creep data.

The raw shear creep data for Sn3.5Ag flip-chip and CCC solder joints is plotted in Figure 20. Some datasets show the expected continuity while others do not. For example:

• The 25°C flip-chip data by Yang, H. et al. shows continuity with the 27°C CCC data by Darveaux et al. However, the same data falls below the 10°C data points by Wiese et al.

• Besides differences in the average stress and strain definitions, which by themselves may lead to discrepancy in the data, differences in the high stress locations and the solder joint shapes may be in part responsible for the differences between the Wiese et al. data and the other two datasets by Yang et al. and Darveaux et al. Also, from past finite element analysis (Lau & Pao, 1997), stress concentration patterns are expected to be quite different for barrel-shaped and hour-glass-shaped flip-chip solder joints. Last, differences in the amount of terminal metals dissolved in solder may contribute to the higher creep resistance of the Darveaux’s CCC and Yang’s flip-chip solder joints.

• Of the three 80°C flip-chip data points by Yang et al., one of them lines up with the Darveaux CCC data at 80°C but the two other data points are over one order of magnitude below. Strangely, these two data points also fall below the Yang’s data points at 25°C.

Figure 21: Fit of Sn3.5Ag flip-chip data to correlation of Darveaux's CCC shear data.

The Wiese et al.'s and Yang’s flip-chip data is shown on the plot of the Darveaux's CCC data correlation in Figure 19. The master-curve of the CCC data is as given by equation (35) that was derived in the previous sub-section using the "Datafit" regression of the original data by Darveaux et al. The 10 X correlation band around the CCC data in Figure 21 is also defined by upper- and a lower-bound lines that are an arbitrary factor 10 = 3.16 times above and below the centerline. The flip-chip data of Wiese et al. and Yang, H. et al. is also shown on the same plot:

• As discussed above, the 25°C flip-chip data by Yang et al. fits rather well within or close to the CCC master curve and correlation band.

• One of the 80°C data points from Yang et al. is close to the correlation band (near its lower bound) whereas the two other 80°C data points are outside and below the band. As pointed out above, it is not clear why this is so.

• The flip-chip data from Wiese et al. is outside the CCC data correlation band and falls within a band that is offset from the CCC master curve by a factor 1.7 X along the stress axis. This indicates that the flip-chip data correlates to the CCC data (within a 1.7 X factor in stress) and follow similar trends in terms of the temperature effect. However, the CCC joints appear to be more creep resistant and to have higher strength. This is possibly due to more strengthening of the CCC joints close to the soldercomponent interface, differences in stress concentration factors for a barrel-shaped versus an hourglass shaped joint or a combination of those two effects.

• The Wiese et al.’s data in Figure 21 also shows a potential "paste / manufacturing" effect. Looking at the shifted master curve and the 10 X correlation band around it, the majority of the "PT" paste data points are below the centerline, whereas most of the "DG" data points are close to or above the centerline.