Effects of SAC alloy copper dissolution rates on PTH processes: cost and performance justify use of certain alternatives to SAC 305/405.
Copper dissolution during PTH rework is not a new phenomenon. In some cases, high degrees of copper dissolution occurred under SnPb conditions when reworking larger, more thermally massive PCBs that required longer contact times (i.e., server type product). In some instances, a "low-melt" alloy was incorporated into the SnPb process to permit longer contact times required for 2X rework. (1,2) This however, was required only on an extremely small percentage of card types. The majority of PTH reworks were capable of using a single alloy type (i.e., eutectic SnPb37). Use of Pb-free alloys such as SAC 305 or 405 inflicts this issue on the majority of product types--regardless of size and complexity. The main reasons for this increase in copper dissolution using SAC 305/405 are the increased operating temperatures, along with the change in the alloy composition itself. Primarily, the increase in tin content (%Sn) causes higher rates of copper dissolution. Aside from incorporating other techniques to attempt to reduce the degree of copper dissolution (reducing contact time using barrel cleaning techniques, flow rate reduction methods), it is clear that a change in Pb-free alloy is the main solution required to establish a safe PTH rework process window, which the industry has been accustomed to when using Sn-Pb alloy.
A handful of alternative Pb-free alloys are commercially available: variations of SnAgCu or SnCu-based alloys with varying degrees of elemental additives such as nickel, germanium, bismuth, antimony and others. These additives add certain properties that help control grain structure, which acts to reduce the degree of copper dissolution and provides other soldering benefits. Table 1 lists specific alloys in this study and their respective melting ranges. The impact of these additives from a metallurgical point of view will be briefly discussed.
To date, previous studies on the copper dissolution phenomenon have resulted in several key outcomes. First, industry will likely be forced away from SAC 305/405 alloys during PTH processes. (3,4) The main reason for this is based on the high copper dissolution rates of these alloys, which have a direct impact during PTH rework. This is a concern, especially for high reliability, long-life products that may require multiple reworks during their lifetime. It has been shown that SAC 305 and 405 alloys have copper dissolution rates two times higher compared to a SnCu + Ni doped alloy and conventional eutectic SnPb. Calculated copper dissolution rates are directly correlated to the allowable PTH rework process window. It can be seen that significant improvements to the allowable process window are made when using an alternative Pb-free alloy that offers lower rates of copper dissolution (Figure 1).
[FIGURE 1 OMITTED]
The second conclusion was that copper dissolution rates will vary from barrel-to-barrel across a connector and will even vary within a single barrel based on geometry (Figure 2 a,b). Specifically, the knee of the PTH barrel will dissolve at a faster rate as compared to the annular ring or barrel wall. (5) This is explained by atomic diffusion theory correlating the effects of surface geometry on the rate of atomic diffusion, as well as proximity to the solder flow. The main reason for the variance across a connector has been linked to flow rate. Typical flow-well design uses a single central opening through which solder can flow. This design point often results in higher dissolution rates within center barrels of a connector rework location. Center pins are often subjected to turbulent solder flow, while edge pins are subjected to laminar flow. Modifications in flow-well design have proven effective in reducing solder turbulence, resulting in reduced copper dissolution rates during solder fountain rework. (6)
All the "alternative Pb-free alloys" included in this study are primarily SnAgCu or SnCu-based alloys with some controlled levels of additives. The additives help control the final grain structure, thus improving final joint appearance, wetting and flow characteristics. Effects include improved barrel fill and reduced copper dissolution. The type, wt% and combination of the base elements and additives have an impact on controlling the aforementioned properties. This section summarizes the impact of each of the common additives, specifically on copper dissolution from a metallurgical point of view.
Copper dissolution kinetics. Before discussing each additive's impact on copper dissolution, the basic mechanisms behind how copper dissolution occurs must be explained.
The copper dissolution process can be considered to be occurring by the following mechanisms: (7)
1 Departure of atoms of the solid surface.
2. Diffusion into the solder melt.
Diffusion-controlled processes result in a uniform attack, while preferential etching of grain boundaries may recognize interface-controlled reactions. In this study, smooth copper/intermetallic interface without any signs of grain boundary attack was detected. The mechanisms themselves occur in series, and the slowest one determines the overall kinetics of the process. The most general dissolution rate equation is: (8)
C = [C.sub.s](1-exp(-K(A/V)t))
Where C is the solute concentration at time t, K is the solution rate constant and V is the volume of liquid. This equation can be applied for diffusion- or interface-controlled processes. The solution rate constant K is D/[DELTA] for the case of diffusion control, where D is the diffusion coefficient in liquid and [DELTA] is the effective concentration boundary layer thickness. In general, the boundary layer thickness is less than 0.1 mm. This boundary layer is a layer of liquid existing immediately adjacent to the solid copper interface/intermetallic layer (Figure 3, online). The copper concentration gradient exists within this layer. During the diffusion-controlled process, the liquid boundary layer, formed during solder fountain rework, is an important feature of copper dissolution.
The liquid diffusion boundary layer thickness is a function of the physical properties, solution velocity and diffusion coefficient. The dissolution rate increases with increasing peripheral velocity, which is relevant to the fountain rework situation. Other influences in reducing copper dissolution would be based on the type of additive present in the bulk solder. The exact influence of each is discussed below.
SnPb, SnAgCu and SnCu-based alloys. It has been found that the tin component of most solders reacts with the copper substrate. (9) In the case of SnPb solders, only the tin components react, since copper is nearly insoluble in liquid lead at soldering temperatures and forms no intermetallic compounds with it. Therefore, the Sn-rich solders dissolve more copper than eutectic SnPb.
With increasing copper concentration in the solder, the rate of dissolution decreases because of the concentration gradient reduction. Thus, solders with 0.7% copper remove less copper from the plating layer than solders with 0.5% copper. Therefore, based on this, the SAC 305 alloy (0.5% copper) copper dissolution rate should be greater than that of the SAC 405 alloy (0.7%Cu). This will be further illustrated in the results section.
Effect of additives. An effect of a reactive third and fourth component within a binary- and ternary-based alloy, respectively, is not properly understood yet and cannot be predicted without experimentation under different conditions. Below is a brief summary of the expected reaction of common additives, specifically with respect to controlling copper dissolution.
Some components such as antimony and bismuth may dilute the Sn-rich solder and reduce copper dissolution in molten solder. Typically these elements improve solder grain structure, strength and/or ductility. However, they are prone to defects such as fillet lifting and contamination issues.
Some metals that react with tin may increase the effective copper concentration in the bulk molten solder and slow the copper dissolution rate by reducing the concentration gradient. The nickel additive substitutes copper in [Cu.sub.6][Sn.sub.5] particles, forming a complex (Cu,Ni)[.sub.6][Sn.sub.5] compound. The ternary [Cu.sub.x][Ni.sub.y][Sn.sub.z] formation was reported as well. (10) For better understanding of the nickel influence on copper dissolution rate, the next level of complexity of the dissolution should be described. At the interface of copper and molten solder, an intermetallic layer is formed that grows at the copper side, and at the same time dissolves at the solder side. (11,12) This intermetallic layer (Figure 3) is very small: about 0.02-0.03 [micro]m. (13) The nickel component is concentrated in the intermetallic layer. In this case, small quantities can have major effect on the concentration gradient within the copper layer. This gradient significantly slows copper dissolution.
[FIGURE 2 OMITTED]
When an additive reacts with dissolving copper, the effective copper concentration is lowered and the dissolution rate consequently raised. The oxygen component may cause the copper dissolution rate to increase when the molten solder is exposed to air. If a species such as germanium (an antioxidant), which reacts with oxygen, is present, the dissolution rate will go down.
The test vehicle (Figure 4, online) selected was a current OEM product card. It was selected because of its thermal nature to characterize copper dissolution on a large, thermally massive PCB. The dimensions are 8.9 x 19.4", and 0.096" thick. It consists of 24 layers, which are 0.5 or 1 oz. copper plating, with multiple ground connections and has an OSP surface finish, with a high Tg FR-4 laminate. The board is populated with 32 inline DIMM connectors. Table 2 (online) shows the copper dissolution DoE matrix. Each alloy's testing was separated into two experimental groupings (Exp.# 1 and 2) with two test vehicles used.
The first experimental grouping (1-1 to 1-4) included three different Pb-free alloys (SnCu + Ni, SAC 305 and SAC 405), along with a eutectic SnPb control cell. The second grouping of alloys (2-1 to 2-4) included four Pb-free alloys, a six-part SnAgCu alloy with Bi + "other" additives, a SnAgCu with an Sb additive, a SnCu alloy with two minor additives (<0.1% each), and a repeat of the SnCu alloy with a Ni and Ge (<0.01%) additive. Each sample was exposed to two different contact times (30 and 50 sec.) using each alloy.
The primary factors varied in the experiments were solder alloy and contact time. All other potential variables were set as constants: flux type, equipment type, preheat method and preheat time. However, as each alloy has a different melting temperature, the pot temperature was changed to ensure constant superheat (Table 3). This is important, as alloy temperature has an impact on dissolution rates. Keeping each alloy's superheat consistent was an attempt to remove the variability caused by each alloy's differing melting temperatures.
In each group, a bare board was used for the experiment, separating individual DIMM locations into the DoE samples (Figure 5, online). The total sample size including all seven alloys was 96 (48 samples per board), with each sample consisting of approximately 14 barrel locations. This would equate to 28 separate copper thickness measurements taken per sample.
Each DoE sample was cross-sectioned along the entire length of the sample at a middle row and copper thickness measurements taken. Copper thickness measurements were focused solely at the knee location (Figure 6). In absence of a current IPC specification for remaining copper plating thickness after rework, an OEM specification of 0.0005" of remaining copper plating was used. (14)
The statistical results indicate that both contact time and alloy type have a significant impact on the copper dissolution rate (Table 4).
The interaction plot (Figure 7) illustrates the copper dissolution results of each alloy at both 30 and 50 sec. exposure times. The values on the y-axis are the measured remaining copper thickness of the knee plating after a 30 and 50 sec. exposure to molten solder (x-axis). The alternative Pb-free alloys have been coded (Alloy A, B, C and D) with the results of SAC 305, 405 and the SnPb control cell identified in the chart.
The results show two of the alternative Pb-free alloys performed the best with respect to copper dissolution rate. Specifically, Alloy C and D results are comparable to those of the SnPb control cell. In addition, Alloy B's result also showed a statistically significant difference from the performance of SAC 305/405 alloys and, although showing poorer results, it had no statistical difference compared to Alloys C, D and the SnPb cell. Although it had a similar slope to the other alternative Pb-free alloys, Alloy A showed a slightly higher copper dissolution occurrence at 30 and 50 sec. exposure times. Alloy A has a statistically similar rate of dissolution compared to the SAC 305/405 alloys. Figure 7 reveals that Alloy A's results drop below the specification limit of 0.0005" of remaining copper plating at approximately 40 sec., which is a concern.
As the Pb-free PTH process window is strongly dependant on total cumulative exposure to solder, the final calculated copper dissolution rates can be correlated to a maximum expected allowable process window for each alloy studied. Figure 8 illustrates each alloy's process window using three different specification limits of remaining copper plating thickness for comparison sake. Also illustrated on the graph is the total time required to 1X (50 sec.) and 2X (100 sec.) rework the DIMM connector assembled on the same test vehicle. Based on this, it can be seen that both Alloy C and D are capable of performing up to a 2X rework using any of the three specification limits listed. Alloy B is capable of performing a 1X rework using any of the three specification limits; however, it falls short of being able to complete a 2X rework. Finally, Alloy A along with SAC 305 and 405 are capable of performing a 1X rework, however, only when using a specification limit of 0.0002" of remaining copper or less. All three are incapable of completing 2X PTH rework.
Effect of ground planes on copper dissolution. Within each sample, every barrel connected to a ground layer was identified (Figure 9--red). This provided the opportunity to determine whether barrel layer construction played a role in the rate of copper dissolution.
The statistical results showed that the layer construction did in fact have an impact on the copper dissolution rates. In fact, the results showed a statistically significant difference between the ground and signal layers (Table 5, online).
Specifically, the results indicated that the barrels attached to a ground layer resulted in a lower occurrence or rate of copper dissolution as compared to barrels attached to a signal plane (Figure 10).
The ground layer attach's impact on copper dissolution rate is at this point not completely understood. This finding, however, potentially can be correlated to results obtained from an earlier study that indicated increasing the preheat temperature of a PCB prior to rework helped reduce the rate of dissolution. (4) The question is: Is this finding simply an interesting fact, or is there any way in which this information can be used to help reduce the impact of copper dissolution at the PTH rework process by incorporating board design changes (i.e. moving or adding ground layers)? Realistically, the effort and cost in altering existing product board design layer construction and PTH components would make this impractical. In addition, this solution also poses a "Catch-22," as the existence of thermal ground planes, although incorporated as a thermal relief for the connector, is the primary cause for requiring increased solder contact time during the PTH rework process to adequately reflow the joint. Therefore, the benefit of the ground connections on reducing copper dissolution may be outweighed by the required increased contact time needed to rework the connector, which is a significant factor in causing copper dissolution. However, this finding is useful in helping pinpoint inspection for copper dissolution during both the visual inspection process after rework, as well as during any destructive lab analysis.
Reliability of SnCu + Ni Alloy vs. SAC 405
This section includes a sample of reliability results on SnCu + Ni and SAC 405 alloys, obtained from a separate, earlier large-scale internal project performed using a test vehicle named RIA2. (15) SnCu + Ni alloy testing was selected as it was, at the time a readily available "alternative Pb-free alloy," which could provide a good basis for comparison against the common SAC 405 alloy in the wave solder process. Within this project, a cell consisted of a number of OSP boards assembled with a variety of PTH connectors and bottomside leaded devices (Table 6 online).
Boards were wave-soldered using both a SnCu + Ni alloy and SAC 405 in an air environment, using a WW flux. Each board was then subjected to accelerated thermal cycling (ATC), from 0-100[degrees]C, with 10 min. dwells (per IPC-9701). As the primary focus of the study was not on PTH technology, nor the bottomside leaded devices, these locations were not in-situ monitored throughout the ATC. Therefore, probability plots could not be created. However, valuable information was derived from analyzing both the time zero joints, as well as the joints after 6000 cycles. These results are discussed below.
Time zero analysis. There were no significant differences in appearance between the SAC 405 and SnCu + Ni joints. For PTH joints, SnCu + Ni solder joints were shiny and SAC 405 joints were rough. The leaded components, however, were shiny using both SAC 405 and SnCu + Ni alloys. In addition, fillet lift-off was observed in the PDIP joints soldered with SAC 405 and SnCu + Ni.
Analysis after 6000 cycles (0-100[degrees]C). In each case, there were no failures due to completely open joints found after 6000 cycles under a 0-100[degrees]C ATC environment. Cracks on the surface were visible under optical microscope, but they did not result in an open joint (Figure 11 a-d). Stress relaxation corner cracks (Figure 12 a, b) and cracks along the pins were formed in both SAC 405 and SnCu + Ni connectors. These cracks started at both the top and bottom fillets and propagated along the pin side. Although both types of joints are robust, the SnCu + Ni alloy is more prone to crack formation under thermal fatigue. In addition to the main cracks, the SnCu + Ni alloy contains many narrow transgranular cracks (Figure 13 a, b). This type of crack is not found in the SAC 405 alloy. Fillet lift-off cracks propagated slightly during the thermal cycling for both the SnCu + Ni and SAC 405 alloys, but did not cause catastrophic failures. Figure 14 shows an example of the fillet lift with SAC 405 alloy. SOIC components did not fail after 6000 cycles at 0[degrees]-100[degrees]C. However, minor deformations and crack nucleation were visible at the heel and toe areas. The damage is more significant in the SnCu + Ni joints than in the SAC 405 (Figure 15 a-d).
Impact to Manufacturing
Changes in the alloy/chemicals will inherently cause an impact to manufacturing. Changes to the alloy used within PTH processes will have a significant impact on quality, reliability, supply chain, processing and cost. Each of these impacts needs to be addressed before making a change. Altering only the Pb-free alloy used within the solder fountain process, or changing both the solder fountain and wave processes, will cause varying degrees of impact to many different aspects within a manufacturing environment, as summarized below.
Quality and reliability. There has been little study of the effects of performing a forced mixed Pb-free PTH rework (i.e., SAC 405 assembled connector with an alternative Pb-free solder fountain alloy). The impact would need to be studied to validate this as a process, specifically, studying the degree of mixing and final joint microstructures formed. In addition, pot contamination levels would need to be closely monitored and corrective rehabilitation actions established. An earlier study looked at "mixed Pb-free" scenario occurring during touchup, with the results showing no detrimental effects from a process quality or reliability standpoint. (16)
Before introducing an alloy into the wave process, its reliability performance requires validation (e.g., ATC and thermal/mechanical testing). This reliability effort could take nine to 12 months, depending on the testing performed.
Each product will need to be requalified using the new alloy to assess the process and final assembly quality if either the solder fountain or wave machine alloy is altered.
Supply chain considerations.
* Supply sourcing issues will need to be addressed.
* New business relationships need to be established.
* Global distribution and support verified.
If an OEM adopts the new alloy for existing products built using SAC-based alloys, EMS and OEM agreements will need to be made on the use of alternate Pb-free alloys. As one EMS typically builds for many OEMs, changes will have to be agreed to by multiple OEMs building within the same factory.
Cost considerations. Many alternative Pb-free alloys cost less than high silver content alloys such as SAC 305/405. Alloys B, C and D studied are each lower in cost than SAC 305/405. Figure 16 illustrates the cost of each alloy studied relative to the cost of SnPb. This would therefore have a positive impact on manufacturing, offering a significant savings, especially when changing the wave process, which typically consists of 2000 lbs. of solder.
However, there are also "switching costs" to consider before changing the Pb-free alloy. Examples of these costs include labor costs in loading in new alloy; costs incurred when swapping SAC alloys for alternatives; and logistics changes on the shop floor, including dedicated wave solder machines with new alternative alloys, pot cleaning, and pot controls.
Process considerations. Having multiple Pb-free alloys used within the manufacturing process will require strict material segregation and monitoring to reduce the chance of contamination.
Until effects of various contamination levels are completely understood, frequent monitoring of solder pot contamination levels would be required when altering the solder fountain alloy only.
The high copper dissolution rates of SAC 305/405 alloys are indeed the main driver in the requirement to change the Pb-free alloy used for solder fountain rework. This change could also potentially drive a change in the alloy used within the wave solder attach process to simplify the manufacturing process and reduce pot contamination during PTH rework. However, the push to replace SAC 305/405 alloy with an alternative Pb-free alloy within the wave soldering process could easily be justified from a financial perspective as well. Some alternative Pb-free alloys offer a 50% reduction in cost compared to SAC 305/405 alloys, which can be quite significant if this saving is shared among multiple wave soldering machines.
This study indicated three "alternative Pb-free alloys" (B, C and D) could be used as potential replacements for SAC 305/405 based on copper dissolution rate results. Each of the three alloys showed statistically similar dissolution rates to that of SnPb. There are, however, other considerations to consider before selecting an appropriate alternative alloy. These include the final joint quality produced and reliability. In addition, supply chain considerations, logistics and costs are important factors in selecting a replacement Pb-free alloy.
Ed.: This article was first published under a different title at the SMTA Pan Pacific Symposium in January 2007 and is republished with permission.
The authors would like to thank each of the solder suppliers involved in supporting this study. The authors would also like to acknowledge the support of IBM, in particular Jim Wilcox and Jim Bielick. In addition, thank you to Celestica's support throughout this study, in particular to Linda Scala, Thilo Sack, Irene Sterian, John McMahon, Marianne Romansky, Heather McCormick and Jeffrey Kennedy.
Craig Hamilton is process development engineer at Celestica (celestica.com), email@example.com. Matthew Kelly is senior engineer at IBM (ibm.com); firstname.lastname@example.org. Polina Snugovsky, Ph.D., is engineering consultant at Celestica; email@example.com.
Ed. For a list of references, please see the online version at circuitsassembly.com.
Craig Hamilton, Matthew Kelly and Polina Snugovsky, Ph.D.
Table 1. Alloy Composition and Melting Range Alloys Melting Range ([degrees]C) SnPb37 SnPb 183 SnAg3.8Cu0.7 SAC405 217-220 SnAg3.0Cu0.5 SAC305 217-224 SnCu0.7Ni0.05Ni(<0.01)Ge1 SnCuNi 227 SnCu0.7Bi-X-Y2 SnAdCu+Bi 217-228 SnAg2.5Cu0.8Sb0.5 SnAgCu+Sb 217-219 SnCu0.7-X-Y SnCu + X 227 1. U.S. Patent #6180055 2. Patent #PCT/GB2005/004609
|Printer friendly Cite/link Email Feedback|
|Comment:||Effects of SAC alloy copper dissolution rates on PTH processes: cost and performance justify use of certain alternatives to SAC 305/405.|
|Date:||Jun 1, 2007|
|Previous Article:||Pb-free wave soldering flux application: popular flux systems are compared for PTH penetration, coverage accuracy, maintainability and chemical...|
|Next Article:||Mass reflow assembly of 01005 resistors: tombstone ratios of co-assembled 0201s and 01005s show the latter are more stable across a variety of...|