ABSTRACT
Minor voiding in Ball Grid Array (BGA) solder joints is a common phenomenon, so these voids are not often considered to pose a vital threat to solder joint reliability. It has been found, however, that incorrect handling conditions can create open solder joints during reflow. This paper investigates some causes of voiding and open solder joints, and recommends ways to avoid them. Various design and assembly parameters have been studied to understand the phenomena. Several factors that affect the formation of voids are reported. The investigation reveals that the cause of void-induced open solder joints is primarily outgassing of moisture from components where proper handling precautions have not been observed. If proper handling precautions are followed, these components are shown to have none of the excessive voiding described.

INTRODUCTION
Surface Mount Technology (SMT) has evolved over the past decade from an art into a science, with the development of well-understood design and assembly guidelines. BGA components especially have been found to be robust and user-friendly devices. These packages merge readily into existing processes and reflow profiles. However, as the solder ball size and pitch decrease, the SMT design and assembly characteristics become more critical in controlling certain types of solder joint defects. Board Level Reliability (BLR) focuses on the complex interaction of various materials under the influence of ambient thermal experienced during the operation of the end equipment. In BGA components, the solder joint itself must be successful in accommodating:

• Cyclical strains due to expansion mismatches
• Warping and transient conditions
• Impact stresses
• Non-linear material properties
• Solder fatigue behavior influenced by:
• Geometry
• Metallurgy
• Stress relaxation phenomenon
• Temperature cycle conditions.

Solder joint voids are generally formed by pockets of gas trapped during the creation of solder connections between the component terminals and the Printed Circuit Board (PCB). Although voids may range widely in size and location, void formation may be classified into four main ^[1] types . Materials, Methods and machines, Environmental and Human factors, In this paper, we investigate some rather unique material and environmental factors affecting voids formed during the reflow process. The basic structure of the component studied in this report ^[2] is shown in Figure 1. It is a molded 0.8mm pitch BGA which uses a single layer metal pattern on polyimide tape. The solder balls are near eutectic. This package, called MicroStar BGA^TM, was developed at Texas Instruments for cost reduction and miniaturization. This component has been successfully surface mounted in high volume for nearly three years. As of July '99, more than 60 million units have been shipped. It is believed the phenomena studied in this paper are unique to this generic type of BGA, rather than specific to MicroStar BGA.

Figure 1. MicroStar BGA^TM; is a trademark of Texas Instruments

MATERIALS AND METHODOLOGY
The experiments covered in this paper fall into two categories:

In most cases, the experimental factors are greatly exaggerated or accelerated compared to normal conditions in order to generate time-zero failures or to shorten the time required for long-term failures to occur.

Underfill was not used in any of these experiments.

Visual inspection of solder joints, which was conducted after one or more reflows, consisted of microscopic cross- sectioning, SEM or transmission x-ray techniques. Image analysis software was not used in this report, so the visual techniques have a only a qualitative value.

The BLR testing is based on non in-situ electrical monitoring of daisy-chain components assembled to a special PCB. Electrical measurements are made in the initial state and then at intervals of 100 temperature cycles. The daisy-chained units are made using the standard assembly process including the silicon chip and gold bond wires. When a daisy-chained package is assembled on the PCB, a complete circuit is formed which allows continuity testing. The circuit includes each solder ball, the metal pattern on the die, the bond wires and the PCB traces.

SOLDER PASTE
Solder paste is recommended when mounting fine pitch BGAs for four basic reasons^[3]:

• It acts as a flux to aid wetting of the solder ball to the PCB land.
• The adhesive properties of the paste tend to hold the component in place during reflow.
• It helps to compensate for minor variations in the planarity of the PCB or solder balls.
• Solder paste contributes to the final volume of solder in the joint and allows the total volume to be optimized.

Paste selection is normally driven by overall assembly requirements. In general, the "no clean" compositions are preferred due to the difficulty in cleaning under the mounted component. Most assembly operations have found that no changes in existing pastes are required by the addition of fine pitch BGAs to a PCB, but due to the large variety of board designs and tolerances it is not possible to say this will be true for any specific application.

Nearly as important as paste selection is stencil design. A proactive approach to stencil design can pay large dividends in assembly yields and lower costs. The typical stencil hole diameter is the same size as the pad area, and 125 to 150um thick stencils have been found to give the best results. Good release and a consistent amount of solder paste and shape are critical, especially as ball pitches decrease. The use of metal squeegee blades, or at least high durometer poly blades, is important in achieving this. Paste viscosity and consistency during screening is another variable that requires control.

REFLOW
Solder reflow conditions are the next critical step in the mounting process. During reflow several things occur in short time span ^[4]:

• The solvent in the solder paste evaporates.
• The flux cleans the metal surfaces.
• The solder particles melt.
• Wetting of the surfaces takes place by wicking of molten solder.
• The solder balls collapse.
• The process is completed with the solidification of the solder into a strong metallurgical bond.

The desired result is a uniform solder structure strongly bonded to both the PCB and the component, with minimal voiding and an even fillet at both ends. Conversely, when all the steps do not carefully fit together, major voids, gaps, uneven joint thickness, discontinuities and insufficient fillet can occur. While the optimum reflow cycle depends on the system and paste composition, there are several key points all successful cycles have in common.

The first of these is a warm-up period sufficient to safely evaporate the solvents. This can be done with a pre-heat or bake, commonly reflected as a hold time in the cycle at evaporation temperatures. If there is less solvent in the paste (such as high viscosity and high metal content paste), then the hold can generally be shorter. However, if the hold time is not long enough solder splatter can occur. Secondly, successful reflows commonly have uniform heating across the component and PCB. Uneven thickness and non-uniform solder joints may be an indication that the profile needs adjustment. There can also be a problem when different sized components are simultaneously reflowed. Care needs to be taken when profiling an oven to be sure that the indicated temperatures are representative of what the worst case components experience. This concern is heightened with infrared (IR) reflow.

In summary, successful reflow cycles strike a balance among temperature, timing and length of cycle. Poor timing may lead to oxidation, excessive voiding, premature paste dry-out, poor wetting and splattering. Some process development is advisable to guarantee a good process window for each paste and reflow combination. Extremely fine tuning of existing profiles are driven by the interplay of such factors as solder paste particle size, flux activity, metal percentage, ramp rates, peak and hold temperatures, board construction and atmosphere.

OTHER REFLOW ISSUES
While fine pitch BGA components offer improved board- level manufacturability, throughput and yield compared to larger components, there are some factors to keep in mind, especially as the ball pitch decreases. One item of special attention is the moisture sensitivity. Manufacturers normally specify the moisture sensitivity level for each type of component and it is important to respect these conditions. The time out of a dry environment should be controlled according to the label on the packing material or other specifications. This will prevent excessive moisture absorption, which can lead to solder joint defects as discussed later in detail. Also, the PCB may twist and bow during reflow. This problem becomes more pronounced as PCBS decrease in thickness. Potential problems from this effect may show up as mis-aligned and mis-formed solder joints, or other discontinuities. Proper support of the PCB through the furnace, balancing the tab attachments to a panel, and even using a weight or fixture to stiffen the PCB may be worth investigating.

EXPERIMENTATION AND RESULTS
In order to investigate thoroughly the effects of exposure to high moisture levels and different types of reflow, several individual experiments were conducted. The component studied, the 64-lead MicroStar BGA, has been qualified by Texas Instruments to JEDEC Moisture Level 3 at both the component and the board level ^[5].See Table 1. The board level reliability testing was conducted by non in-situ electrical monitoririg of daisy-chain resistance components, as shown in Figure 2, mounted to a daisy- chain PCB design as represented in Figure 3. Electrical measurements made in the initial state and then at intervals of 100 temperature cycles. The daisy chained units are made using the standard assembly process including the silicon chip and gold bond wires. When a daisy-chained package

is assembled on the PCB, a complete circuit is formed which allows continuity testing. The circuit includes each solder ball, the metal pattern on the die, the bond wires and the PCB traces.

Table 2 contains a summary of the materials, conditions and results from our study of various reflow conditions. The samples were tested with electrical and visual inspection. The results were as expected for conventional reflow systems. The vapor phase (VP) reflow system here is used as a reference, since it is not found in most assembly houses. In the case of IR reflow systems; failures were detected in low levels. IR reflow furnaces

are rarely found in production today, although many customers use IR reflow for prototyping or one-shot assemblies.

OBSERVATIONS
A micro-cross section photo representative of the open solder joints found in this test is shown in Figure 4. It can be seen that the solder ball does not completely de-wet from the component, but simply separates, leaving a wetted solder surface. The break occurs within the bulk of the solder joint near the "neck". This is the area where the solder ball joins the component through an opening in the tape. This experiment indicates that the combination of IR reflow and high moisture level exposure can create low levels of open solder joints.

On the rare occasion that an IR system is used, it is likely to be in a prototype or engineering development environment. In this type of situation, moisture content is more difficult to control since the components are usually exposed to engineering time lags not found in production. In this respect, the use of IR reflow is problematic unless special attention paid to controlling the moisture level per the manufacturer's specification.

ANALYSIS
Although this phenomenon is the result of improper handling conditions, it is desirable to find the root cause so as to understand the pre-cursors and possible side effects. To this end, other inspection techniques were used to narrow-in on the area of the open solder joint.

Figure 5 shows a high magnification cross section in the area of a failing joint. A detailed examination of these results and photos indicate the probable root cause to be outgassing of moisture within the package during reflow. The mechanisms appear to be as follows:

• If the moisture sensitivity recommendation has not been respected, excessive moisture absorbed into the package can vaporize, and seek an escape pathway to the outside of the package.
• The moisture is vaporized under pressure, so the temperature at which this occurs is somewhere above 10OC. Empirical evidence suggests that this occurs when the solder is in the molten state. In some cases, the vapor escape pathway can lead to a solder ball connection on the package.
• As the high-temp vapor enters the component solder via, it may escape from the via. For example, down the sidewall, leaving the solder joint intact.
• Alternative y, the vapor may create an open solder j oint by displacing molten solder in the joint. This seems to occur under certain conditions:
  • The solder is softened to some degree such that the vapor may create a break in the joint.
  • After the break, as the solder reflows, the two separated solder surfaces do not combine to reform the original joint. This could be due to a lack of solder volume (if solder has been spat away from the joint), or due to oxidation effects on the separated surfaces. A diagram illustrating this mechanism is shown in Figure 6.

FOLLOW-UP EXPERIMENTS
Additional experiments were conducted to corroborate the analysis suggested above. The focus in this area was to identify some measurable factors that should follow from the premise and to verify these.

As mentioned earlier, we expected that the grade of the seal formed by the via opening and the molten solder would depend to a large extent upon the geometry. One easily measurable factor in this system is the initial solder ball diameter. With this in mind, we investigated the effect of initial solder ball size on open solder joints in the uncontrolled conditions described above.

The results may be found in Table 3. As seen, the largest solder ball size did increase the level of opens. As expected, a larger volume of solder covers the via opening more completely, thus providing a stronger seal against the molten solder collapsing back into the via. In this table, the .60mm diameter solder ball clearly shows some effect, In addition, a small solder ball will permit the escape of vapor around the ball without displacing

solder and disrupting the joint. A large solder ball tends to trap escaping moisture, thus increasing the tendency of the vapor to displace solder in the via area, creating the open joint. However, it is not advisable to select a precise optimum solder ball size from this specific experiment for three reasons:

• The diameter of the via openings vary among components. This table is based solely on a via opening of .375mm diameter.
• Since only four different solder ball sizes were examined, the resolution is not great enough to select an optimum solder ball size.
• Finally, the thickness of solder paste on the PCB varies among applications. The solder in this paste contributes to the overall solder volume and will change the solder ball size during reflow.

Another area investigated was the moisture characteristics of the materials in the component. The results from these measurements are listed in Table 4, The materials were measured in amounts proportional to those found in the overall component This clearly indicates that a relatively greater amount of moisture is absorbed by the substrate materials and the chip attach material. This supports the root cause analysis presented earlier, since the prominent vaporized moisture pathway appears near these interfaces.

CONCLUSIONS
The precise extent to which the individual effects described above occur, and the extent to which they are related, is not known in a quantitative way. However we have drawn some general conclusions based on the empirical evidence and supporting opinions.

• Investigations conducted on fine pitch BGAs, including the MicroStar BGA, have demonstrated 100% board assembly yield and acceptable thermal cycling results even with some inconsistencies in the solder paste process^[6].
• No open solder joints were observed when the proper handling precautions were followed. Always adhere to the manufacturer's specification regarding the JEDEC moisture sensitivity level.
• IR reflow is clearly more sensitive to this phenomenon. When using IR reflow, special attention must be paid to controlling the moisture level.
• An extra large volume of the solder in the joint during reflow tends to increase the number of open solder joints under these inappropriate conditions. However, the proper amount of solder volume varies among components. In this case, the initial solder ball should be less than 0.60mm diameter.

ACKNOWLEDGEMENTS
The authors wish to thank the following colleagues who performed crucial research and provided valuable information used in this report:

REFERENCES
1. Anthony Primavera, Roland Sturm, S waminath Prasad and K. Srihari, Factors That Affect Void Formation In BGA Assembly, Proceedings of Institute for Interconnecting and Packaging Electronic Circuits, October 1998.
2. K. Ano et al., MicroStar BGAs, Texas Instruments Application Note, TI SCJ 2328, July 1998.
3. Gerald Capwell, High Density Design with MicroStar BGAs, Texas Instruments Application Note, SPRA47 1A, July 1998.
4. Charles Williams et al., MicroStar BGA Packaging Reference Guide, Texas Instruments Application Note, SSYZ015, December 1998.
5. Mohamed K. E1-Ghor, Mark Peterson, Kumar Pavuluri, Jimmy Settles, Paul Hundt, James Berry, and Hirotoshi Takeda, Evaluation Of 64 Pin MicroStar BGA Package, Proceedings of Fourth Annual Pan Pacific Microelectronics Symposium, February, 1999.
6. Puligandia Viswanadham, Steve Dunford and Ted Carper. Board Level Reliability Evaluations of 40, 32 and 30 Mil Pitch Ball Grid Array Packages Over -40 To 125��� C, Proceedings of Fourth Annual Pan Pacific Microelectronics Symposium, February, 1999.