Selecting solder mask material: how to avoid falling in the cracks.

Laminate chip carrier packages continue to expand and mature in the semiconductor industry. In addition, increasing customer demand for quality and reliability requires that closer attention be given to the material component of the laminate substrates.

The goal of the study described in this article was to establish the key solder mask criteria that ultimately affect and improve solder mask performance with respect to cracking. To accomplish this goal in substrate form and also isolate the solder mask performance, we created a method of stressing the solder mask without damaging the core material or its components carrier.

A material stress analyzer (MSA, patent pending) was developed to allow us to measure the solder mask performance under various stresses and processing conditions by employing elements of standard three-point bend methodologies, among other analytical techniques. This analytical device was extremely effective in both visual and quantitative differentiation of both materials and processing conditions.

The MSA (Figure 1) was designed to quantitatively and qualitatively measure the effects of stress/strain relationships on solder mask to resultant cracking. Discovered early in the study, the sensitivity required to adequately measure marginal stress/strain relationships was well beyond simple three-point bend test apparatus capability. Therefore, a much more sophisticated system was developed to measure the slightest stress/strain under virtually any processing conditions. The system has the following merits:

[FIGURE 1 OMITTED]

* Applies the combination of visual and stress data.

* Deflection and visual data reflect the fundamental characteristic of solder mask proper ties such as modulus, film strength and film elongation.

* Condition on the surface of the material need not necessarily be related to direct stress/strain curve measured by classic stress/strain test method.

* The end point of interest is the condition of the surface, which can only be determined by visual analysis first and then followed by stress/strain correlation.

* Allows useful testing on multilayer and composite materials that cannot be achieved through traditional bend procedure and stress/strain correlation.

* Allows snapshots of different stages of crack mechanisms.

Experimental Methods, Procedures

Solder mask was coated onto 200 micron foil that was pretreated with an adhesion promoter. The promoter used was an electrochemical product similar to that currently being used by most substrate suppliers in Asia. The laminate material samples were then subjected to varying solder mask conditions to test the ability to withstand cracking under various stress conditions. Observations of the fractures were done at 40X magnification. Listed below are the tests and Design of Experiments (DOE) conditions examined in this study.

Test vehicles

* solder mask film over substrate

* solder mask film over copper foil substrate

* solder mask film over unclad/clad laminates

* solder mask film over inner core materials

Solder mask materials

* Supplier I: A, B, C, D, E

* Supplier II: F

* Supplier III: G

Substrates

* 200 micron copper foil

* BT laminate

* epoxy FR-4

Response variables

* measure deflection values at crack initiation point

* capture visual images of crack mechanism at different stages of bend

Factors

* degree of thermal cure exposure

* post bake vs. no post bake

* thin film vs. thick film

* degree of ultraviolet (UV) bump exposure

* resin to hardener mixed ratio

Effects of laminate

* diagonal stress vs. stress along glass weave

* actual routing configurations

* different types of resin systems

Experimental Results

The fracture characteristic of the different solder mask materials yielded interesting results. First and foremost, solder mask A is most brittle. Cracks propagated easily and became worse with an increasing bend radius. Adhesive failure or delamination around crack lines could be observed at maximum deflection on solder mask A. Initiation of crack(s) occurred at approximately 0.8 mm to 1.0 mm (Figure 2).

[FIGURE 2 OMITTED]

Initial crack formation of solder mask B, C and D occurred at a similar degree of bend as solder mask A. However, the cracks depicted in Figure 3 indicate that solder mask B, C, D and E materials exhibited less brittleness with less crack propagation as compared to solder mask A. The crack mechanism of solder mask B, C, D and E consisted of finer and smaller fractures in the form of numerous interconnecting networks of hairline cracks. The surface condition appeared to be less damaged at a higher degree of bend with no adhesive failure or delamination detected at maximum deflection on these solder mask materials. Solder mask F had stronger film strength prior to cracks but appeared brittle, resembling solder mask A in crack propagation behavior at a higher level of stress/strain region.

[FIGURE 3 OMITTED]

Solder mask F fracture initiation (Figure 4) occurred at about 1.5 to 2 times more deflection than solder mask A to E. Solder mask E fracture behavior was somewhere between the performance of solder mask A and solder mask B, C and D. The initiation of cracks occurred earliest around 0.6 mm to 0.8 mm of strain for solder mask E, but the fractures were much finer with short propagation and harder to detect.

[FIGURE 4 OMITTED]

Solder mask G material exhibited a much higher degree of ductility than solder mask material A to E Most interestingly, no fracture was observed at all levels of bending with solder mask G (Figure 5).

[FIGURE 5 OMITTED]

The results of the testing were broken down into five primary categories:

1. effects of resin to hardener mixed ratio

2. effects of thermal cure

3. effects of solder mask film thickness

4. effects of UV cure

5. relationship of laminate material to cracking.

Results Analysis

Figures 6 and 7 illustrate the effect of processing parameters on solder mask cracking.

[FIGURES 6-7 OMITTED]

The effects of resin to hardener mixed ratio had the opposite effect on solder mask F and solder mask A as compared to solder mask B, C and D. The material became more brittle as the hardener content ratio went up in the case of solder mask A and E At 50% hardener content, the resistance to fracture decreased by approximately 35% on solder mask F and by 20% on solder mask A, accompanied by more severe long fractures. On the other hand, dramatic improvement to crack resistance occurred at the 50% hardener content in the case of solder mask B, C, and D.

No fracture could be observed on solder mask B and C at the 50:50 mixed ratio (Figure 7). Slight fractures could still be seen on solder mask D under the same condition. Deflection values of solder mask E were the lowest to initiate cracks, and the fracture characteristics were not altered as significantly as the other solder mask materials by different levels of resin to hardener mixed ratios. Crack appearance became slightly more pronounced at higher levels of hardener on solder mask E material.

[FIGURE 7 OMITTED]

The effects of thermal cure conditions on solder mask crack performance were also found to be significant. As shown in Figure 6, fracture initiation for solder mask F and solder mask A at thermal cure of 285[degrees]F at 45 minutes occurred with 25% and 15% lower deflection than at 300[degrees]F at one hour, respectively. These solder mask materials were found to have optimum crack resistance at thermal cure condition of 300[degrees]F with a one-hour bake. An additional post-thermal cure for one hour at 300[degrees]F did not improve the crack resistance property of any of the solder mask materials tested.

As shown in Figure 7, additional post bake process increased solder mask brittleness and decreased the deflection values. Solder mask B and C, which did not exhibit any crack at 1:1 mixed ratio, showed some cracks at 1.6 mm deflection after post cure. Therefore, under-curing and over-curing conditions were not desirable conditions.

The effect of varying solder mask thickness was found to be statistically insignificant. The DOE compared 30 micron and 60 micron film thickness on solder mask A and E No change was indicated in the initial crack formation and crack mechanism on either solder mask material. No other solder mask materials were tested under these thickness conditions. Fracture tendency should be expected to increase with thicker solder mask film coating. The same held true for variance in UV cure conditions. We looked closely at the UV cure conditions of 0 mJ, 500 mJ and 1500 mJ. Again, no change was indicated in crack resistance on any solder mask materials.

Finally, when looking at the relationship of laminate materials to the performance of solder mask resistance to crack, changes in the solder mask processing parameters alone might not have sufficient impact on crack resistance. Other methods such as the MIT flexibility test, acoustic emission analyzer were incorporated with the material stress analyzer to demonstrate the impact of laminate core material on solder mask cracking. The laminate core material was found to have as much effect on solder mask cracking as the solder mask itself. Resin type, filler type and the number of glass plies used in the laminate construction were shown to significantly affect the nature of laminate and solder mask cracks. Due to time and depth of study, these findings were not included in this report.

Predictive Modeling/ Quality Control

Based upon simple three-point bend test applications, we were able to derive an equation that demonstrates predictive capability (Figure 8).

[FIGURE 8 OMITTED]

Derivation of the equation translates the deflection data from the test sample of a certain size to other lengths. The amount of deflection required to crack solder mask film is proportional to the square of the sample length. As shown by the equation, a smaller amount of bend would be required to initiate crack, as the sample size gets smaller. Correlation of deflection value to sample length could be used as a tool to predict crack failures related to bending and handling tolerances of a particular manufacturing process such as drilling, singulation and plating.

Also, application of the three-point bend test and visual analysis could be used as a quality assurance tool to troubleshoot solder mask anomalies by comparing the deflection values and visual information to a known standard.

Experimental Results Conclusion

This study's goal was to determine the best-known methods for solder mask processing, the design optimization for solder mask and a predictive model for solder mask. The summary of our finding includes the following:

Identify solder mask best known method (BKM)

Solder mask A was found to be most brittle and least resistant to cracking. Solder mask F could withstand more bending before developing fractures, but, after initiation of cracks, they were as pronounced as the cracks in solder mask A. Solder mask B, C, D and E exhibited more ductility with crack mechanism that consisted of numerous networks of fine fractures that did not worsen as much as solder mask A at higher bend. Solder mask G had the best crack resistant properties with no observable fractures during the bend test. Solder mask A and F became more brittle as the hardener component of the mixture increased, but the opposite effect was observed on solder mask B, C, and D. No crack was seen at 1:1 resin to hardener mixed ratio on solder mask B and C. An additional post-bake process would not improve the crack resistance property of the materials. Under-curing or over-curing would degrade solder mask resistance to cracking.

Design optimization for solder mask and laminates

With the addition of a stress sensor to the three-point bend mode material stress analyzer, developing a correlation of solder mask and laminate cracks as a function of stress, strain, modulus, thickness, coefficient of thermal expansion (CTE), laminate build-up construction and layout configuration might be possible.

Predictive model for manufacturing processes related to cracking

The mathematical correlation between deflection and sample size could be used to identify processing issues such as handling and bending tolerance for a particular process.

Summary

By utilizing the material stress analyzer, we were able to confirm the best-known methods of solder mask processing conditions. We created a predictive model to calculate bend stress required to initiate micro cracks in the solder mask as well as laminate material.

Between the material stress analyzer and acoustic emission analyzer, the exact moment of material separation can be determined in the laminate or solder mask material. The MSA, however, through visual depiction, will offer a better solution for those looking for the exact moment of stress/strain to initiate the crack in either the solder mask or laminate material. Using the MIT test as well allowed the confirmation of the solder mask effects on copper traces and laminates.

From the testing performed, the newer version of solder mask materials B, C, and D clearly performed much better with resistance to cracking than the earlier version of solder mask A. Solder mask G performed the best through our evaluation, but moisture resistance test (MRT) performance is unconfirmed.

Acknowledgement

The authors would like to thank Ed Holdsworth for his extensive work and effort in developing the material stress analyzer for this application.

Ron Huemoeller [[phone] ext. 5153, e-mail: [e-mail]] is vice president of substrate technology and development, and Sukianto Rusli [[phone] ext. 5132, e-mail: sukir@amkor. com] is senior engineering manager of emerging substrate technologies, both with Amkor Technology, Chandler. AZ.