Corey Reichman

Interface failure in lead free solder joints

The phenomenon of interface failure in lead free solder joints was explored using solder joint array tensile testing. The effects of pad metallization, solder alloy, reflow conditions, and post reflow thermal aging were quantified. The joint strength ranged from 5 to 115MPa. The joint ductility dropped to zero in some cases. The interface microstructure and failure mode were characterized for each combination of factors. Most of the trends were linked to microstructural features of the interface. A ductile-to-brittle transition strain rate (DTBTSR) was defined as a metric to quantify the performance of a specific joint relative to interface failure. The DTBTSR ranged from 10-3 /sec to 10/sec for the conditions studied

Mechanical Properties of Lead-Free Solders

Mechanical testing of solder joint arrays was used to characterize lead-free solder alloys under both shear and tensile loading. The alloys studied included Sn4.0Ag0.5Cu, Sn3.5Ag, Sn0.7Cu, Sn3.0Ag0.5Cu, and Sn1.2Ag0.5Cu0.05Ni. All of the lead-free alloys exhibited time and temperature dependent creep behavior over the entire temperature range from -55 C to 125 C. A SINH creep law was effective in describing the steady state creep behavior over the entire range of conditions. Lead-free alloy strength increases with increasing Ag content. The relative strength of tin-lead versus lead-free alloys depends on the strain rate and temperature regime. At high strain rates and low temperatures, eutectic tin-lead is the strongest alloy. At low strain rates and high temperatures, eutectic tin-lead is the lowest strength alloy. Simulated stress-strain hysteresis loops for both accelerated test and field use conditions showed dramatic differences between the various alloys. Higher Ag content results in a larger stress range but a smaller strain range during temperature cycling.

Ductile-to-brittle transition strain rate

Solder joint failure mode depends on factors such as loading mode (shear vs. tensile), temperature, and strain rate. For any given loading mode and temperature, there is a transition from ductile failure to brittle failure as strain rate is increased. This ductile-to-brittle transition strain rate (DTBTSR) is one of the best indicators for robustness relative to impact loading. A higher value for DTBTSR is better. DTBTSR was measured for a wide range of lead free solder joints. The effects of pad metallization, solder alloy, reflow conditions, thermal aging, test temperature, and loading mode were characterized. DTBTSR was found to improve with 1) electroplated Ni/Au over other finishes (as long as the substrate supplier has a well controlled process), 2) less reflow time above liquidus for Cu pad metallization, 3) mild thermal aging, 4) higher test temperature, and 5) shear loading over tensile loading.

Effect of joint size and pad metallization on solder mechanical properties

Metallurgical analysis, mechanical testing, and finite element analysis were conducted to understand the effect of joint size and pad metallization on solder behavior. A wide range of design and material variables was evaluated. The pad metallization affected both the intermetallic compounds at the joint interfaces and those dispersed in the bulk solder. NiAu pad metallization resulted in more creep resistant joints than Cu. These effects were more pronounced at lower test temperatures. Solder joint creep resistance increased with joint size. Larger joints were also more prone to brittle interface failure than smaller joints. This was true in both package level tests and board level tests. Finite element analysis indicated that the predicted fatigue life in cyclic drop or temperature cycle testing can be significantly affected by the test vehicle used to generate data for constitutive constants. More creep resistant behavior resulted in lower strain and work for both temperature cycle and drop test conditions. When comparing strain vs. work as a damage indicator, it is seen that work is less sensitive to variations in the constitutive constants.

Crack initiation and growth in WLCSP solder joints

Solder fatigue crack initiation and growth was studied in WLCSP assemblies. Samples were surface mounted to test boards, then extracted at regular intervals of thermal cycling. Crack lengths in the corner solder joints were measured using dye-and-pry technique. Four temperature cycle conditions and six solder ball alloys were evaluated. For all alloys and conditions, the cracks initiated near the component pad interface and propagated down into the solder bulk somewhat. Sn0.7Cu had some propensity toward cracks that propagated at approximately 45 degrees from the pad. 63Sn37Pb and Sn0.7Cu had much higher crack growth rates on the outboard side of the joints compared to the inboard side. SAC405, SAC305, SAC125Ni, and Sn3.5Ag had more similar growth rates on outboard and inboard sides of the joints. These differences are believed to be due to the temperature dependence of the creep deformation and damage accumulation in each alloy. 63Sn37Pb, SAC125Ni, and Sn0.7Cu showed a greater number of cycles to crack initiation compared to the other alloys. This trend is likely due to a lower creep resistance, which resulted in reduced tensile stresses at the edge of the solder joints. Sn0.7Cu and SAC405 had the lowest crack growth rates under most of the temperature cycle conditions. The estimated cycles to failure based on the present crack initiation and growth data was compared to previous data from like assemblies that were electrically monitored. It was found that the average crack growth data from individual joints over-estimates the mean fatigue life of components. However, the average + 3 a crack growth data provides a good estimate of first failure fatigue life from components. Statistical arguments are proposed to explain these results.

Field use environment for a FCBGA in a laptop computer application

A laptop computer was disassembled, and several thermocouples were attached to the die, heat sink, substrate, and motherboard. The laptop was then re-assembled and run under various conditions to measure the effects on temperature distribution. The assembly was then deconstructed and samples were extracted for material property measurements. Elastic modulus and thermal expansivity were measured for the heat sink, substrate, motherboard, thermal interface pad, and underfill materials. Typical temperature rise of the IGP die above ambient was 40C. Video use increased the temperature by 5C to 10C. Wrapping the laptop to constrict airflow increased the temperature by 15C. Hence, the operating temperature range is approximately 55C to 80C. The substrate and motherboard are hotter on the side facing the CPU. There are gradients of up to 20C in the structure (difference between hottest and coolest regions). The initial temperature change rate during a power cycle is 5C/sec for the IGP die.

Solidification behavior of lead free and tin lead solder bumps

An optical observation method was used to characterize the solidification behavior of WLCSP and Flip Chip solder bumps on actual product devices. The WLCSP solder ball diameter was 350um, and Flip Chip diameters were 110um and 80um. Sn3.5Ag, Sn2.3Ag, SAC305, Sn0.7Cu, and 63Sn37Pb alloys were investigated. Cooling rates were varied between 0.10C/sec and 3.9C/sec. In addition, Vickers Hardness testing was employed to look for any correlation between undercooling and mechanical response of the alloy. Solidification was found to occur 4C to 60C below the solidus temperature (undercooling = 4C to 60C). The range of undercooling for the bumps on a given sample was between 8C and 40C. The sequence of solidification was quite random. Cooling rate had a minor effect on undercooling behavior for rates between 0.5C/sec and 3.9C/sec. For the very slowest rate of 0.1C/sec, the amount of undercooling decreased. 63Sn37Pb and 96.5Sn3.5Ag had the least amount of undercooling and the tightest range of undercooling. 96.5Sn3.0Ag0.5Cu and 99.3Sn0.7Cu had the most undercooling and the widest range of undercooling. Flip chip bumps had 5C to 10C more undercooling than WLCSP balls. There was no correlation between undercooling and Vickers hardness, within the range of conditions studied here. There was a slight increase in hardness for SAC305 samples with a faster cooling rate during solidification. Finally, understanding of the solidification behavior in Flip Chip joints was used to interpret a specific reliability test result.