Under testing, uniform joint microstructures and reliable joints were observed.

A novel low melt rework solution was previously reported by the authors from the results of a multi-phase Stage 1 investigation.⁶ The work included the evaluation of three commercially available alloys and the creation of optimized rework profiles with reduced rework temperatures. Based on the metallurgical analysis and a limited ATC study on RIA2 test vehicles, a low melting In-containing alloy was selected for further investigation.

A comprehensive thermal fatigue reliability study of the solder joints formed after reworking BGAs was performed in Stage 2 testing and analysis. This work was performed on RIA3 test vehicles using the low melting indium-containing alloy selected in Stage 1.

Background (Stage 1)

The initial low melt rework investigation was performed in four sequential phases on RIA2 test vehicles. The first phase was an investigation of joint microstructures resulting after rework using three different low melt alloys. Three commercially available alloys were selected (Table 1). Alloys A and C were alloys containing indium, and Alloy B was a Bi-containing alloy.


In this phase, all BGAs on a limited number of RIA2 boards with OSP finish were reworked in air. Hot gas rework profiles were created and optimized for each solder paste. The criteria for rework parameter optimization were proper shape of solder joint after rework, minimized voiding, uniform microstructure and absence of or reduced low-melt phases in the solder joint.

In the second phase, optimized rework profiles were used. This phase was performed on a limited number of RIA2 boards with an ENIG finish. Only alloys A and B were included for this part of the investigation. Alloy C was dropped off the matrix due to the unknown properties of the high In-content solder joints.⁵

The third phase compared 1X and 2X rework processes in air and nitrogen environments. In this phase, boards with immersion Ag, OSP and ENIG finishes were reworked using alloy A. The influence of rework on adjacent components, PBGA 196 and CSP 46 (Figure 1) was also investigated at this stage.


Two scenarios were analyzed. First, both PBGA 196 and CSP 46 were removed at the same time. Then the CSP 46 component was replaced using the low melt In-containing alloy A. After that, the PBGA 196 was placed using the same low melt alloy. The second scenario was more thermally challenging. The CSP 46 component was removed and a new component was placed using low melt In-containing alloy A. Then the PBGA 196, which was originally assembled with SAC 385 solder paste, was removed using high heat to melt the SAC alloy. Finally, the new PBGA 196 was placed using low melt In-containing alloy A.

In the fourth phase, the influence on thermal fatigue life of each of 1X and 2X reworked components using optimized profiles was studied on limited RIA2 boards using alloy A and alloy B. On all boards, two identical BGAs were assembled. During this phase, one of each component was reworked using low melting alloys and one remained untouched (primary attach). After low melt rework, the boards were tested using accelerated thermal cycling (ATC). Results indicated solder joints reworked using In-containing alloy A performed better or comparable to primary attach joints. Joints reworked with Bi-containing alloy (alloy B) failed before the non-reworked component; therefore, alloy B was excluded from further analysis.

Experimental (Stage 2)

Based on results obtained in Stage 1, In-containing alloy A was chosen for further analysis in Stage 2. In Stage 2, a comprehensive investigation of the thermal fatigue reliability of the low melt reworked solder joints was performed per IPC-9701.1 BGAs on RIA3 boards were reworked using In-containing alloy A. The reworked solder joint performance was then compared to that of the primary attach joints, as well as to components conventionally reworked using SAC 385 paste.

The RIA3 test vehicle was 8" x 10" with 12 copper layers and available in 0.093" and 0.125" thicknesses. This TV was designed to represent a mid-range complexity product with greater assembly process challenges, and was made from a Pb-free compatible laminate material capable of withstanding the higher temperature requirements of lead-free processing. Surface finishes used on the RIA3 test vehicles for this investigation were immersion silver (ImAg) and electrolytic nickel immersion gold (ENIG). This test vehicle is daisy-chained and permits four wire in-situ monitoring of the components during ATC (Figure 2).

The RIA3 TV covers a range of component technologies (Table 2). In the TV design, two PBGA 196 components were added at the bottom side for the rework cells, which were a mirror image of the two PBGA 196s populated on the topside.


Primary SMT assembly was performed with noclean SAC 385 paste using a standard 10 zone reflow oven, in air.

One-time hot gas rework was performed on all BGAs on RIA3 TVs with different thicknesses. Site re-dressing was performed following component removal. Low melt solder (In-containing alloy A) was then applied to the re-dressed sites for all components. The rework process was performed under nitrogen for all boards. Figure 3 shows locations of the reworked components. The matrix for this investigation is in Table 3.


Cell 6-1 and 6-2 were designed to investigate the thermal fatigue reliability of solder joints formed on all components reworked using low melt In-containing alloy A on 0.093" boards with ImmAg and ENIG finishes. Cell 6-3 and 6-4 were included to assess similar properties, but on 0.125"-thick boards. These low melt rework cells 6-1 to 6-4 were incorporated in a much larger experimental matrix in Celestica’s RIA3 Lead-Free project. This project included conventional rework as well. A comparison between low melting and conventional rework was also performed.

Test Protocols

Thermal shock preconditioning was completed for all test boards prior to the start of accelerated thermal cycling. This preconditioning was intended to simulate the worst-case thermal conditions a product might experience during shipping from the assembly site to the customer. Cards were electrically tested before and after the preconditioning. Nets showing a 20% or greater resistance increase were deemed to have failed. After completion of this test, preconditioned cards were loaded into the thermal chambers for further cycling. ATC was then carried out according to IPC-9701 for standard area array packaging. The profile selected was 0° to 100°C, with a minimum dwell time of 10 min., measured on the solder joints. The full test duration was planned to be 6000 cycles. Cards were mounted in racks, which held the cards in position inside the chambers and permitted air to circulate freely around them (Figure 4). These cards were in-situ monitored for the duration of the test using dataloggers. Failure was defined as five consecutive measurements showing a 20% or greater increase over the initial reference value. The initial reference value is the electrical resistance of a daisy net during the hot dwell on the first thermal cycle.

Microstructure of Reworked Solder Joints

As-assembled and reworked solder joints were examined before and after thermal cycling using optical and scanning electron microscopy and x-ray analysis. Differential scanning calorimetry (DSC) analysis was also performed to study the solder joint solidification after mixing of the low melt solder with SAC 385 balls during rework. The DSC results showed, if full mixing was achieved, the low melt liquid was completely consumed and the resulting composition crystallized at a reasonably high temperature range.

In general, in this study, the low melt solder paste was melted at temperatures below the SAC 385 solder ball melting point. The solid solder balls dissolved in the molten solder. In general, the dissolution process depends on the rework parameters of temperature and time, and may cause full or partial mixing. In addition to temperature and time, the ratio between solder ball and solder paste volume is especially important for complete mixing, and causes different microstructures for different components. Snugovsky, et al described the theory of solid solder ball dissolution in molten solder.⁷

Metallurgical analysis was performed on all solder joints after rework, and their microstructure was characterized. No time-zero defects or open joints were found using a BGA scope, optical microscopy or SEM. Solder joints were properly formed, collapsed normally and well-shaped (Figure 5). It was also observed that for all reworked joints, the microstructure was uniform and fully mixed. There was no portion of initial solder ball visible in the cross-sections of reworked solder joints. This finding confirms the component solder balls were fully dissolved in the liquefied solder paste during reflow, and the resulting liquid solidified during cooling.

Transmissive x-ray was used to inspect the assemblies for defects and to assess the level of voiding. No assembly defects were noted, but all cells showed some voiding. In all cases, voiding was within acceptable levels per IPC-A-610-D.⁹

The DSC heating curves for reworked CBGA and PBGA196 solder joints using In-containing alloy A are illustrated in Figure 6. These curves confirm the metallurgical observations reported above.

As shown in Figure 6(a), for reworked CBGAs, the melting of a joint starts at 208°C, 17°C higher than the In-containing solder paste melting point. It melts in a relatively narrow temperature range, and stops melting at 212°C. This range is much below 217°C, the melting temperature of SAC solder balls. Reworked PBGA 196 solder joints finish melting at an even lower temperature (209°C), confirming SAC 385 solder balls are completely consumed, Figure 6(b). Melting begins at 178°C.

The final composition of CBGA joints after rework using In-containing alloy A is 3.7-3.8 % Ag, 3.8-4.0 % In, and 0.4-1.0 % Cu, depending on the board surface finish. The microstructures of these joints are very similar to those of pure Sn-Ag-Cu alloys. CBGA reworked solder joints have Ag3Sn plates, Sn dendrites, and eutectic in inter-dendritic spaces. The number of Ag3Sn plates is small, and they are not large in size, Figure 7(a). The EDX analysis revealed indium in the Ag3Sn particles and the Sn matrix. These data are consistent with the literature on SnAgIn alloy microstructure formation.8 The ternary liquidus projection phase diagram⁹ (Figure 8) may be used to interpret the microstructures. It shows that for low In-content (0 to approximately 4 atom %), the compound phase is the Ag3Sn type, with some indium substituted for tin. The matrix phase is the βSn type.

PBGA 196 joints after rework contain 3.6-3.7 % Ag, 6.0-6.2 % In, and 0.4 -1.0 % Cu. Tin dendrites and eutectic are present in the microstructure. Large primary intermetallic particles that have a specific flower shape, shown in Figure 7(b), are detected in the PBGA 196 reworked solder joints. This compound contains a significant amount of indium and is more likely the Ag2.7 (In,Sn) type. It was found9 that for intermediate In content in alloys (approximately 4 to 6 at %), the compound phase has the approximate composition of Ag2.7 (In,Sn), where the indium and tin contents are approximately equal. It also may be the Ag2In type that forms in alloys with a high In-content (greater than approximately 6 at %). The occurrence and size of the particles depend on the rework parameters, and may be significantly reduced by optimizing the profile.

The reworked CSP46 microstructure is similar to PBGA 196. There are no significant differences between 1X and 2X reworks on ImAg and ENIG finishes. The influence of rework of the adjacent PBGA 196 was negligible.

The reaction intermetallic layer at the board side is (Cu,Ni)6Sn5 with 3-5 % Ni. The intermetallic on the component side is a ternary compound, SnCuNi, with a nickel content in the range of 10-15%. For the OSP and ImAg finished boards, the thickness of the intermetallic layer on the board side is about 5 µm in CBGA joints and 6µm in PBGA 196 joints, in both cases after 1X and 2X rework. The intermetallic layers on the board side of the ENIG boards is much thinner, about 1.8 µm and 2.5µm in CBGA and PBGA 196 joints, respectively.

Both CBGA and PBGA 196 solder joint compositions after rework with In-containing alloy A provided an excellent balance of the desirable properties: strength, plasticity, fatigue life and superb fatigue resistance. The described microstructures are also favorable for high reliability of the reworked solder joints.

There were no differences found between the microstructures of components reworked in air and nitrogen.

Accelerated Thermal Cycling

The ATC results reported below are based on 2500 cycles completed to date. The test is currently ongoing to a target of 6000 cycles.

Weibull plots were generated for CBGA and PBGA196 components reworked using the low melt In-containing alloy A, as well as for primary attached components and components reworked using conventional rework process (SAC 385 alloy). These are the only two components that have failed up to this cycle count.

CBGA.  Plots for CBGAs comparing the thermal fatigue reliability of low melt reworked CBGAs to that of the primary attached parts is shown in Figure 9.

As can be seen in Figure 9, cell 6-1, the thermal fatigue reliability of the solder joints formed after rework using low melt In-containing alloy A is significantly better than that of the primary attached joints where SAC 385 alloy was used. A similar trend was observed for the other cells.

Figure 10 compares the Weibull plots of reworked solder joints formed on thick (0.125”) boards with different surface finishes. This illustrates CBGAs reworked using the low melt In-containing alloy performed slightly better on boards with ENIG than on ImAg.

Figure 11 shows the performance of low melt reworked CBGAs on boards with different thicknesses. No difference was observed between cells 6-1 and 6-3, which had board thicknesses of 0.093" and 0.125", respectively.

Weibull plots for all four low melt reworked cells are shown in Figure 12. The thermal fatigue reliability of CBGA components reworked using low melt In-containing alloy is very similar for all cells. This plot confirms the performance of the ENIG finish on both board thicknesses is slightly better than that of the ImAg finish.

Figure 13 compares thermal fatigue reliability of low melt reworked CBGAs on thick boards (0.125") with ImAg and ENIG finishes with that of the conventionally reworked components on 0.125"-thick boards with OSP finish. The conventional rework was part of the original matrix and was performed on the same test vehicle using the optimized rework process parameters.
 
The surface finish on the conventionally reworked cell was OSP. These plots illustrate the performance of low melt reworked components in both cells 6-3 and 6-4 is comparable to that of the conventionally reworked parts.

In general, CBGAs reworked using the In-containing alloy (with a final resultant joint composition of 3.7-3.8 % Ag, 3.8-4.0 % In, 0.4-1.0 % Cu, and remainder Sn) outperformed non-reworked pure SAC385 joints.

Figure 14 shows the 1% failures for reworked and primary cells 6-1 and 6-2. The 1% failure for the low melt reworked components is better than that of the primary attach. Reworked solder joints outperform as-assembled joints formed on ImmAg and ENIG by a factor of 2 and 6, respectively.

Microstructure after ATC. CBGAs that failed early during ATC were cut from the boards and analyzed using electrical probing (just after retrieval from the ATC chamber), optical and scanning electron microscopy, and energy dispersive x-ray. Microstructures of mixed CBGA joints after ATC were characterized. The analyses revealed the mode for all failures was normal and was identified as thermal fatigue cracks through the solder balls at the component level for all (Figure 15a). In general, for all cells after ATC, CBGA cracks initiated from the solder/component interface. In addition to major cracks that caused open circuits, some board side cracks were also observed, which are normal in the case of CBGAs (Figure 15b). Some diagonal and vertical cracks also formed at the corner balls. These cracks then propagated through the bulk of the solder (Figure 16).

These observations confirm previously reported results for the failure mode of CBGAs.¹¹

PBGA 196. Weibull plots for PBGA 196 components reworked using In-containing alloy A were generated and compared to that of the primary attached (pure SAC 385) for all cells (Figure 17).

Again, as was shown for CBGAs, thermal fatigue reliability of the solder joints formed after reworking PBGA196s using low melt In-containing alloy A is either comparable to or slightly better than that of the primary attached (where SAC 385 alloy was used).

Figure 18 shows thermal fatigue reliability of the low melt reworked PBGA 196 and conventionally reworked components. These data were generated on boards with 0.125" thicknesses. It is shown the performance of the low melt reworked components in cell 6-3, ImAg finish, is significantly better than that of the cell 5-2, OSP finish, particularly in terms of early failures.

Microstructure after ATC. Failure analysis of the low melt reworked PBGA 196 components indicated,  in all cases, failure modes were normal and cracks occurred at the component side (Figure 19).
 
PBGA256, PBGA676 and CSP64s. Event plots for PBGA256, PBGA676 and CSP64 components reworked using In-containing alloy A were generated. Thermal fatigue reliability performance of the primary attached (pure SAC 385) components for all cells were compared to that of the low melt reworked components, and the conventionally reworked components (Cell 5-2).

Figure 20 compares thermal fatigue reliability of the low melt reworked PBGA 256s to that of the primary attached on cells 6-1 and 6-2. It clearly can be seen performance of the low melt reworked components is either comparable or slightly better than primary attached on boards with different surface finishes.

Figure 21 compares performance of the low melt reworked PBGA 256 components to that of the conventionally reworked (5-2). These reworks were all performed on 0.125"-thick boards. Surface finish was ImAg on cell 6-1 and HTOSP on cell 5-2.

Thermal fatigue reliability performance is comparable in these cases.

Figure 22 illustrates thermal fatigue reliability performance of the low melt reworked PBGA676 components compared to that of the primary attach on cells 6-1 and 6-2.

Figure 23 compares performance of the low melt reworked PBGA 676 components to that of the conventionally reworked. These reworks were all performed on 0.125"-thick boards. In all cases, thermal fatigue reliability performance of the low melt reworked components is either comparable or significantly better than that of the conventionally reworked.

Figure 24 compares thermal fatigue reliability of the low melt reworked CSP64s to that of the primary attached on cells 6-1 and 6-2. The low melt reworked components are again showing significantly better performance on both surface finishes.

Microstructure after ATC. Failure analysis of the low melt reworked PBGA 676, PBGA 256 and CSP 64 components indicated, in all cases, failure modes were normal and cracks occurred at the component side (Figure 25).

Conclusions

• The alloy of this low melt rework process contains indium and has a melting temperature close to that of SnPb eutectic. When the low melt solder paste melts, dissolution of the SAC balls occurs. When the rework process parameters such as time and temperature are properly controlled, the result is full mixing of the solder ball with the solder paste, forming a uniform joint microstructure.
  
  
• The reduced temperature of the process prevents component overheating, reduces risk of board warpage and pad cratering, and prevents neighboring and mirror-imaged components from thermal damage.
  
  
• In general, thermal fatigue reliability of low melt reworked components is comparable to or slightly better than that of the as-assembled and conventionally reworked components.
  
  
• There is not a statistically significant difference between the thermal fatigue reliability of low melt reworked components on different surface finishes and on boards with different thicknesses.
  
  
• Reworking BGAs using the low melting alloy provides an excellent combination of processability and reliability.
  
  
• In-containing alloy A, in combination with modified process parameters such as solder paste volume and reflow profile, may be recommended for manufacturing rework and field failure rework. Use of this alloy results in high-reliability solder joints without damage to boards, packages and adjacent components.

 
Acknowledgments
The authors would like to thank Joel Trudell of Celestica for rework profiling and Russell Brush of Celestica for ATC testing and data analysis.

References

1. IPC, “IPC-9701, Performance Test Methods and Qualification Requirements for Solder Mount Attachments,” January 2002.
2. S. Bagheri, et al, “Low Melting Alloys: A New Solution to Lead Free Rework Process Issues,” SMTAI, October 2007.
3. A. Gowda, K. Srihari and A. Primavera, “Challenges in Lead-Free Rework,” Pan Pacific Microelectronics Symposium, January 2002.
4. A. Gowda, K. Srihari and A. Primavera, “Lead Free Rework Process for Chip Scale Packages,” white paper, July 2004.
5. J. Bath, et al, “Lead-Free and Tin-Lead Rework. Development Activities within the NEMI Lead-free Assembly and Rework Project,” SMTAI, September 2004.
6. P. Snugovsky, S.Bagheri, Z. Bagheri and M. Romansky, “The New Pb-Free Assembly Rework Solution Using Low Melting Alloys,” Apex, February 2007.
7. P. Snugovsky, A.R. Zbrzezny, M. Kelly and M. Romansky, “Theory and Practice of Lead-free BGA Assembly Using Sn-Pb Solder,” International Conference on Lead Free Soldering, CMAP Conference, May 2005.
8. IPC, “IPC-A-610-D, Acceptability of Electronic Assemblies,” February 2005.
9. L. Snugovsky, P. Snugovsky, D. D. Perovic and J. W. Rutter “Formation of Microstructure in Ag-In-Sn Solder Alloys,” Journal of Materials Science and Technology, vol. 23, no. 4, 2007.
10. X.J. Liu, Y. Inohana, Y. Takatu, I. Ohnuma, R. Kainuma, K. Ishida, Z. Moser, W. Gasior and J. Pstrus, Journal of Electrical Materials, vol. 31, p. 1139, 2002.
11. M. Cole, M. Kelly, M. Interrante, G. Martin, C. Bergeron, M. Farooq and M. Hoffmeyer, S. Bagheri, P. Snugovsky, Z. Bagheri and M. Romansky, “Reliability Study and Solder Joint Microstructure of Various SnAgCu Ceramic Ball Grid Array (CBGA) Geometries and Alloys,” SMTAI, September 2006.

Ed.: This paper was originally presented at SMTA Toronto 2008 and is published with permission.

Simin Bagheri, Polina Snugovsky, Zohreh Bagheri, Craig Hamilton and Heather McCormick are with Celestica (celestica.com); sbagheri@celestica.com.

• Prev
• Next