Paste chemistry plays a role in printing performance, but there are more questions than answers.

There are two distinct ways to address second-level electrical connection for flip chips: bumping the wafer or bumping the substrate. The former is the traditional mainstream approach, while the latter is a lower-cost alternative.

There are several technologies for substrate bumping: evaporation, electroplating, solder paste deposition and others. Evaporation is well documented and highly reliable, with extremely fine-pitch capability. Yet, it is considerably expensive, relatively slow and capital-intensive. The more common processes are electroplating and solder paste deposition. Electroplating is a complex process that involves several steps that can be considered environmentally unfriendly. The solder paste deposition is a simpler, faster, cheaper and environmentally friendly process. In this process, a metal mask with an opening for the pad area is used to deposit solder paste. Solder paste is printed onto the substrate and reflowed once. Because of overall faster throughput, solder paste printing is lower in cost compared to electroplating. While electroplating offers finer pitch capability, solder paste printing offers material flexibility, which presents advantages when considering Pb-free packages. Because of the varying plating rates required for the Pb-free solders that are typically ternary alloys, electroplating has difficulty controlling compositions. In general, ternary alloys are a greater challenge because they require more complex equipment and photo-resist processing. Solder paste printing, on the other hand, offers greater composition control and material flexibility to support Pb-free alloys.

Background. Though there are several advantages to the solder paste printing process, there are also challenges and limitations associated with printing. Namely, these are printing solder paste deposits large enough to produce bumps with the required height and coplanarity within the constraints of the pitches. The difficulty lies with the stencil manufacturing technologies and solder paste limitations. When printing full area array designs, slumping of overprinted paste results in solder robbing during reflow. Therefore, it is important to use a paste and stencil technology that promotes easy paste release and has adequate slump resistance during print and reflow. In this study, performance of two different paste chemistries is evaluated.

Critical factors to achieving successful bumping using solder paste printing are:

• Printing technique.
• Stencil design.
• Paste formulation.
• Reflow process.
• Substrate flatness.

The current study focuses on the printing part of the bumping process.

Test vehicle. The test vehicle used for the study was a double-sided, 0.012" thick FR-4 substrate, with 0.5 oz copper and immersion nickel gold pad finish. The substrate was designed to include extreme cases (extremely small pad size and web size) to understand the limitation of the stencil printing capability. Pad size varied from 200 to 65 µm, and the pitch/web was also varied to understand the effects of the above factors on the printing capability (Table 1). For better data management, each group of pads was assigned a location number, the top left group being LN1 and the bottom right being LN36.


Each location contained an array of 25 x 25 pads. Some arrays were repeated on the board to understand the printing characteristic as a function of location. To keep the scope of the initial experiment manageable, only designated areas were inspected for transfer efficiency data.

Stencil design. It is well known stencil design is a critical factor affecting the transfer efficiency of any printing process. Print volume and consistency for small deposit could be maximized by carefully choosing the stencil design parameters. A 35 µm-thick electroform stencil was chosen. The aperture size and pitch ratio to the land pattern was 1:1. Stencil apertures were designed as squares to maximize the amount of paste deposited per site.

Paste types. Pastes used in this study consisted of type 6 powder with water-based and no-clean chemistry.

Experimental

An MPM Accela printer was used in this study. Because of the thin substrate, a dedicated tool with an external venture pump was used to hold the substrate flat during printing. During the experiment, the Rheopum chamber pressure was held constant at 1.0 psi. Several boards were printed to stabilize the chamber pressure before the actual DoE was run. To eliminate paste simmering, the stencil was cleaned with solvent and a vacuum after each print.

Print DoE. The experiment consisted of three factors, two levels and full factorial design. The response for the print experiment was wet paste volume and height. A Koh Young KY300 solder paste inspection system was used to characterize print quality, paste volume and paste height. Optical images of the board were collected to obtain qualitative data.

To minimize noise, the experiment was fully randomized with four repeats (four boards per run order). Noise was further controlled by printing all four boards with one print stroke only, front-to-rear squeegee direction.

Several challenges were encountered, the biggest being the SPI’s limited ability to inspect deposits below 100 µm. Therefore, data from pads less than 100 µm were excluded from the analysis.

Gage R&R. To ensure minimum experimental variability, a Gage R&R study was conducted on the SPI by inspecting one printed board 20 times. Volume and height repeatability were well within the acceptable range. Based on this result, the gage was considered adequate.

Results and Discussion

A “repeat” noise strategy was adopted for this experiment to address run-to-run variations. The board was designed to understand the effect of various pad size, web size and pad location. To simplify the discussion, results from two locations are presented here.

Figures 1 and 2 show the main effect plot and Pareto chart for no-clean paste for location 4 (pad size 200 µm) and location 12 (pad size 125 µm), respectively. The two locations behave very differently. As the pad size decreases, complex interactions become very important for the printing process. This is evidenced by the three-way interaction for the 125 µm pad size. Figure 3 shows results for the water-based paste for location 4 (175 µm pad). The main effect is different from no-clean paste, and interaction plays an important role in the TE.


Based on the DoE analysis, an optimum factor combination was determined for each type of paste. Pad sizes smaller than 100 µm showed acceptable paste transfer (qualitatively), but registration was very poor (paste was printed off the pads) because of board stretch. Deposits were well formed, with no bridging or insufficients for 200 µm pad size. As pad size decreases, even with higher area ratio, paste transfer deteriorates because of the extremely small aperture opening. Figure 4 shows the effect of pad size and web on the overall transfer efficiency for the no-clean and water-based paste.


Paste chemistry effect. Paste chemistry appears to have a significant effect on the paste transfer characteristic. No-clean paste provides a more consistent paste transfer, as would be predicted by the aperture size. However, water-soluble paste shows inconsistent paste transfer for 200 and 175 µm pad sizes. Further investigation is necessary to fully understand the phenomena exhibited here.

Pad/aperture size effect. As one would expect, larger pad size provides better transfer efficiency, regardless of the area ratio. As mentioned, the inconsistency of water-based paste for the 200 µm pad requires further investigation to confirm the finding.

Web size effect. Web size shows no effect on paste transfer efficiency. The critical factor to watch here would be the stencil lifespan. As web size gets smaller, the potential for stencil damage is higher.

Ed.: This column is adapted from a presentation at IMAPS 2008. November 2008.

Rita Mohanty, Ph.D., is director advanced development at Speedline Technologies (speedlinetech.com); rmohanty@speedlinetech.com.