Fluxes Effective in Suppressing Non-Wet-Open at BGA Assembly

With the advancement in miniaturization, the die is getting thinner and the solder bump is getting smaller for BGAs. Consequently, the thermal warpage is getting more severe due to coefficient of thermal expansion mismatch between die and molding compound, and often resulted in non-wet-open (NWO) at BGA assembly. NWO has been ailing the industry for a long time, and costly rework is required to remove the problem. Mechanism of NWO has been investigated, and the defect has been attributed to solder paste being stick to the solder bump during development of component warpage. When the temperature reaches melting temperature of solder, the solder paste body coalesced with solder bump body, and the final solder bump becomes a fat bump. Upon cooling, the thermal warpage lessened, but this fat and short bump failed to reach solder pad, and eventually resulted in NWO. Attempts have been made to develop specialty solder pastes to address this issue, with only very limited success. This challenge actually drives the industry look into low temperature soldering in order to reduce the thermal warpage. Unfortunately this path is accompanied with highly compromised solder joint reliability. In this work, a cold-welding barrier method has been developed for suppressing NWO, and will be discussed below.

The mechanism of NWO has been investigated, and the defect has been attributed to solder paste being stuck to the solder bump during progressing of component thermal warpage. When the temperature reaches melting temperature of solder, the solder paste body coalesced with solder bump body, and the final solder bump becomes a fat bump. Upon cooling, the thermal warpage lessened, but this fat and short bump failed to reach solder pad, and eventually resulted in NWO. Figure 1 shows a typical SMT reflow profile with various stages of NWO defect formation.

Figure 1 also shows the similarities and differences between the Head-on-Pillow (HOP) and NWO defects [1]. The main mechanism of NWO is the lifting of the solder paste from the PCB lands at preheat stage. This preheat stage coincide with the early stage of component warpage development. On the other hand, HOP's main mechanism is a gap formed between the ball and paste during reflow when the dynamic warpage is being developed.

In this work, a dummy BGA component with 81 SAC305 balls was used, as shown in Figure 2. This dummy BGA exhibits a ball height of 0.78mm, and a pitch of 4.5mm.

A photo of test Board used in this study is shown in Figure 3, with a board thickness of 1.1mm, a Cu pad thickness of 140μm, a surface finish of OSP, a pitch of 4.5mm, and a pad diameter of 0.6mm.

The soldering materials used in this study is shown in Table 1.

Flux A is a specialty low activity flux developed to reduce NWO defects. Flux B is a creamy flux with regular fluxing activity. Dippable Paste B is solder paste made using Flux B. Two solder pastes (Paste C and D) were selected for testing; they were halogen-free no-clean type paste.

1. Print paste onto test board using a stencil of 100 μm thickness, aperture diameter 0.6mm.

2. The solder bump of dummy BGA is dipped into creamy flux or dippable paste B of a given thickness. This dummy BGA is then placed onto the paste C or paste D printed and followed by reflow.

Flux can be applied to the solder bump surface as a solid coating. The coating methods are shown in Figure 4. The BGA is dipped into creamy flux, and then placed onto a tray. There is an opening at the bottom of the tray allowing the bumps to be suspended in the air without touching the ground. The tray with fluxed BGA is then placed in a baking oven set at 150°C and baked for 5 minutes.

The BGA packaging house can produce BGA with solder bumps coated with solid flux using the above method. Then the BGA coated with solid flux on bumps can be assembled with conventional process. That is, the coated BGA is placed onto paste printed, then followed by reflow.

BGA is prepared with the solder bumps of dummy BGA with or without being dipped in flux or coated with solid flux. Paste was printed onto test board. The dummy BGA was placed onto the board pads with paste, and the dummy BGA/test board sandwich is transferred to a baking oven and baked for 8 min at 180 °C. The dummy BGA board is then pulled off the board after cooling down to room temperature, and the test board pad is inspected under microscope. A soldering material system which results in pads with ⩾70% of paste being retained on the pads was categorized to be good at suppressing NWO defect [1]. The soldering material performed well in baking test rendered joints with no NWO defect, as shown in Figure 6. Figure 6 showed the cross-sectional view of normal BGA joints, joints with NWO defect, and joints assembled with the use of specialty fluxes.

In order to investigate the impact of flux thickness, Flux A is selected to do NWO test. The NWO defects results are shown in Table 2 for dipped with the flux thickness about 25%, 50% and 75% of the bump height (0.78mm). Based on the market feedback, Paste C was knowns to have a high NWO defects and Paste D had a low NWO defects. Paste C and Paste D were compared with conventional assembly. Using BGA dipped with creamy flux could reduce NWO defect significantly. The more flux pickup by BGA bumps, the less the NWO defects resulted, as shown in Table 2.

The weight of flux pickup is shown in Table 3. For the same film thickness of flux, the amount of flux pickup was fairly uniform.

Relation between NWO defect rate and amount of flux pickup at dipping is shown in Figure 7. For both Paste C and Paste D, the NWO defects decrease with increasing flux pickup. In combination with the practical application of industry, we choose 50% of the bump height for more testing.

Paste C has more NWO defects than Paste D. The test results for the two pastes are shown in Table 4 and Table 5, respectively. BGA bumps dipped with flux A or coated with dried flux A can reduce NWO defects significantly. Dipped with paste is similar to a little overprint and does not help in suppressing NWO defects.

Cold welding often happens between solder in the absence of oxide layer, through solid-state diffusion of solder atoms. NWO is caused by solder paste sticking to solder bump by solder powder cold welding to solder bump, thus got lifted from pad when warpage happening. Therefore, if the solder paste can be prevented from sticking to solder bump, NWO can be avoided.

As noted above, a new method has been developed to reduce NWO defects [2]. A flux barrier is set between solder paste and solder bump. At reflow, the flux barrier will prevent solder powder from cold weld to the solder bump, consequently will impede the solder paste sticking to the solder bump, thus effectively eliminate the NWO defects.

The flux barrier must meet the following two conditions: a) suppress the solder paste sticking to the solder bump; b) does not impede the formation of solder joints at reflow. Although both low activity Flux A and medium activity Flux B showed good effect in suppressing NWO, as shown in Table 4 and 5, the low activity flux is considered to have a wider reflow window than the medium activity flux, since low activity flux will remove the solder oxide layer slower, thus is expected to result in further delay before solder powder could cold weld to the solder bump.

Both no clean flux or water-soluble flux are helpful in reducing NWO defects. But if assemble BGA with no clean paste, no clean flux should be selected. Water-soluble flux should be used with Water-soluble paste.

As shown in Table 4 and 5, dipping in solder paste did not reduce NWO, since the powder in dipping paste can cold-weld to both top solder bump and bottom printed solder paste. Accordingly, solder paste couldn't be used as a barrier. It is similar to a little overprint and does not help in suppressing NWO defects.

In this study, solid flux coating on the solder bump was conducted by drying the pickup of creamy Flux A. As shown in Table 4 and Table 5, the effect of solid flux coating on suppressing NWO is very comparable with dippable Flux A. Apparently, a solid flux barrier is expected to be at least equally effective as creamy flux in impeding paste cold welding onto bump, as verified by the data.

As cold welding barrier, both no-clean flux and water soluble flux are considered physically effective. However, a water soluble flux should not be used together with a no-clean solder paste based on compatibility and reliability consideration. Furthermore, only no-clean solid flux is recommended for solder bump coating purpose, mainly based on moisture pickup issue associated with water soluble flux.

With the advancement in miniaturization, the die is getting thinner and the solder bump is getting smaller for BGAs. Consequently, the thermal warpage is getting more severe due to coefficient of thermal expansion mismatch between die and molding compound, and often resulted in non-wet-open (NWO) at BGA assembly. NWO has been ailing the industry for a long time, and costly rework is required to remove the problem. In this study, a “cold-welding barrier” method has been developed in suppressing NWO. At BGA assembly, after printing solder paste onto PCB, the BGA bumps are dipped into creamy flux prior to placing onto paste printed. This flux pickup will effectively suppress NWO by serving as cold-welding barrier. A low activity flux is considered having as wider reflow profile window than a medium activity flux. Alternatively, the solder bump of BGA can be pre-coated with solid flux at packaging house. The effect of solid-flux coating is equally effective in suppressing NWO. As cold welding barrier, both no-clean flux and water soluble flux are considered physically effective. However, a water soluble flux should not be used together with a no-clean solder paste based on compatibility and reliability consideration. Furthermore, only no-clean solid flux is recommended for solder bump coating purpose, mainly based on moisture pickup issue associated with water soluble flux.