Assessing the reliability of soldered joints in lead-free electronic assemblies

There has been much discussion about the merits of the various lead-free alloys available to the electronics manufacturing industry, with their physical and electrical properties being examined and sometimes disputed in great detail. Whatever the end-user's preference, one vital truth remains: the reliability of a lead-free soldered joint should be at least as good as that achieved with tin-lead solders. How does a soldered joint fail? To establish how reliable a lead-free solder is likely to be in service, the factors that affect reliability should first be examined. Perhaps the best starting point is to consider the mechanism by which a soldered joint fails. The typical failure mode for a sound solder joint is Low Cycle Fatigue (LCF), whereby differential thermal expansion through temperature cycling or power cycling continually stresses the joint until the point of failure. Although LCF failures may have a root cause that stems from a flaw in the material - brittle intermetallics or excessive voiding, for example - it should be borne in mind that even a sound joint has a finite life, and can be expected to fail after a certain number of cycles. Reliability is measured as cycles to failure. Bulk alloy properties - a useful guideline? One might reasonably assume the bulk properties of a solder alloy to be a good indicator of its behaviour and reliability in the form of a soldered joint. If this were the case, lead-free assemblies should exhibit higher reliability for two reasons. Firstly, the higher melting point suggests that any stress generated during service is less likely to approach the alloy's yield strength, which appears to be confirmed by the better Low Cycle Fatigue properties of a lead-free alloy in bulk form - Sn-3.5Ag-Cu alloy has been found to be superior by a factor of ten when tested against Sn63-Pb at 25C. Secondly, lead-free alloys possess higher strength, particularly creep strength. But the reality is somewhat different when soldered assemblies are tested. Despite the lead-free alloy's superior creep strength, it shows similar reliability to Sn63-Pb and Sn62-Pb-2Ag under fatigue testing. Conversely, whereas the addition of bismuth degrades lead-free performance under isothermal testing of the bulk alloy, the opposite is seen when a board assembled using Sn-3.5Ag-Bi alloy is used. Clearly the isothermal bulk alloy tests are a poor guide to assembly reliability. Design and process factors for lead-free To see the full picture, a wider selection of factors that affect joint reliability must come under scrutiny. These can be grouped into two main categories - design factors and process factors. The design of the assembly includes such considerations as the size and composition of the components and PCB, and how the board is populated. These details have a bearing on the compliance of the assembly (lower compliance for LCCCs and chip components than for QFPs, for instance) and determine the pattern of stress applied to individual joints. Process factors - variables such as alloy composition, reflow and cooling profiles, degree of oxidation of component termination and board finish, solder paste characteristics and stencil design will influence joint geometry and integrity, and determine the likelihood of excessive void formation or defects occurring at the solder/substrate interface. The microstructure of the joint may also be modified by the dissolution effect of the solder alloy on the board and component finishes. Reliability tests and their interpretation In assessing and comparing the results of reliability tests, it is essential that the test procedure and conditions are taken into account, if misleading or contradictory reliability claims through inappropriate comparison of data are to be avoided. In particular, the criteria that define failure must be carefully chosen. Typical tests include: electrical monitoring, failure being deemed to occur when a specified resistance is exceeded; measurement of joint strength - chip component joints at, say, 50 percent of their original strength and QFP leads at zero percent could be assessed as failures; and microsectioning joints to monitor crack development. Real life: how lead-free compares Having established the factors that affect reliability, reviewing some of the acknowledged test regimes and their results starts to provide a real-life picture of lead-free reliability. First, the reliability testing conditions used in the IDEALS (Improved Design Life and Environmentally Aware Manufacturing of Electronic Assemblies by Lead-Free Solder) programme, which are summarised as: - Thermal shock: -25/+125C, 3 minutes dwell; -20C/+100C and -40/+125C, 30 minutes dwell; - Power cycling: 20-110C; - Vibration: random 50-2000Hz oscillation, 6-43g acceleration, duration 15 minutes, along x, y and z axes; - Failure parameters: measure joint strength of 1206 resistors and microsection failed joints. The 1206 component outline was chosen as it was inexpensive and available with lead-free terminations. The substrate finish used was OSP copper. The results show that the shear strength of 96SC alloy under temperature cycling is very similar to that of Sn62 and Sn63 alloys, reducing by around 50 percent over 3,000 cycles. Power cycling shows a similar trend, the shear strength of 96SC remaining slightly and consistently below that of Sn62 and reducing by about 45 percent over 3,000 cycles. When sectioned, the failure mode for the lead-free solder alloy is virtually identical to that for tin-lead, the fracture initiating in the thin layer of solder beneath the component and propagating through the main bulk of the solder fillet. Next, the NCMS (National Center for Manufacturing Sciences) test regime and results: - Temperature cycling: -55C/+160C, 10 minute dwells, 10?C/minute ramp (test components 20-lead LCCC, 1206, 0805, 80-lead UTQFP, on OSP FR4 PC); - Temperature cycling : -40C/+125C, 15 minute dwells, 15 minute ramps (test components fleXBGA, PBGA, on OSP FR4 PC); - Temperature cycling : 0C/+100C, 5 minute dwells, 10 minute ramps (test components fleXBGA, PBGA, on OSP FR4 PC); - Failure criteria: continuous electrical monitoring - intermittent open circuit (200 ohm). Looking at the LCCC results for number of cycles to first failure at -55C/+125C, Sn43-Bi57 alloy stands out as offering the best performance (at close to its melting point of 138C), while the other lead-free and Sn-Pb alloys exhibit similar behaviour to each other. For the 20-lead LCCCs at -55C/+160C, the target is to match lead-free performance to Sn-Pb performance at -40C/+125C. The bismuth-containing alloys Sn-Ag3.4-Cu1-Bi3.3, Sn-Ag3.3-Bi4.8 and Sn-Ag4.6-Cu1.6-Sb1-Bi1 come closest, but still fall well short of the target, while the best of the SAC alloys manage less than 50 percent of the Sn-Pb performance. For FleXBGA at -40C/+125C, the story is different again, with the SAC and Sn-Ag-In alloys narrowly taking the top spots, but without significant advantage over other lead-frees and Sn-Pb. Under more moderate conditions of 0C/+100C, bismuth-containing alloys again come to the fore, Sn-Ag3.4-Cu1-Bi3.3 and Sn-Ag4.6-Cu1.6-Sb1-Bi1 outperforming Sn63-Pb by almost 300 percent and the best of the SAC alloys by 25 percent. Accelerated life testing vs. real life These findings raise the question of the effects of regime severity upon the results. Comparing just two of the alloys, upon normalisation the trend is more apparent (see Fig. 3): at 0C/+100C, Sn63-Pb exhibits a fatigue life only half that of Sn-Ag3.5; at -40C/+125C this rises to around 85 percent; and at -55C/+160C, it leaps to around 220 percent. Conclusions and the way forward Confusing and contradictory as these results may at first appear, the following observations may prove helpful in their interpretation: - Bulk alloy properties, including fatigue and joint strengths, are not a meaningful guide to lead-free reliability performance - High-tin, lead-free solders containing silver, copper, antimony and bismuth show broadly similar reliability to Sn63-Pb37 and Sn62-Pb36-Ag2 solders - Different Sn-Ag-Cu alloys possess similar reliability characteristics - Sn-Pb solders exhibit a higher fatigue life than lead-free solders under more extreme conditions (high temperature and strain ranges) - Under less extreme conditions (lower temperature and strain ranges), lead-free solders display up to double the fatigue life of Sn-Pb alloys - Accelerated testing at high strain ranges extrapolates differently to less extreme conditions for lead-free and tin-lead alloys - Addition of up to five percent bismuth may improve reliability - provided there is no lead contamination that could cause the formation of a low-melting Sn-Pb-Bi phase - As yet, there is no sign of a fatigue-resistant solder, regardless of alloying additions - Solders with melting points close to peak temperature during cycling seem to perform particularly well, e.g. Sn43-Bi (mpt 138C) at -55C/+125C and Sn63-Pb (mpt 183C) at 150C-160C peak. Research continues as we approach the July 2006 deadline for lead-free implementation, but for now, the most reliable choice while lead can potentially occur anywhere in the assembly is still one of the SAC alloys. In the future, when lead is successfully eliminated from board.