Author: Liang Yin
Abstract: The performance of solder joints in an accelerated thermal cycling (ATC) test depends on the accumulation of fatigue damage, which is affected by various simultaneously evolving microstructure features. A mechanistic understanding of microstructure evolution and damage accumulation in Pb-free solder joints will help us better interpret thermal cycling test results. In this study various components, including Ball Grid array (BGA), Land Grid Array (LGA), Thin Small-Outline Package (TSOP), Quad Flat No-Lead (QFN) and surface mount resistors, were compared for their failure mechanisms. The focus was on the evolution of microstructure, particularly recrystallization, and its correlation to fatigue crack propagation. A general understanding of damage accumulation was proposed, based upon the observed recrystallization behavior and the fatigue life dependence on ATC parameters.
Author: Babak Arfaei
Abstract: Effects of solder alloy, volume and pad finishes on various aspects of microstructure and the corresponding thermomechanical properties of SnAgCu solder joints were investigated. Particular attention was focused on the behavior of solder joints with interlaced Sn grain morphologies. Crossed polarizer microscopy and scanning electron microscopy (SEM) were used to characterize Sn grain structures. Precipitate sizes and distributions were measured using backscattered scanning electron microscopy and quantified using image analysis software. Mechanical properties including hardness and indentation creep were measured. Results show that the amount and frequency of interlacing increased as the joint size decreased, as the amount of Ag in the solder increased, and if the joint was reflowed on ENIG substrates. The interlaced structure was harder and more creep resistant compared to the common beach ball morphology. Image analysis results showed this to be related to much higher densities of secondary precipitates in the interlaced regions. A mechanistic understanding of the microstructure is discussed and recommendations are made as to the design of more reliable solder joints.
Author: Babak Arfaei
Abstract: We have recently demonstrated a significantly longer life in accelerated thermal cycling for Land Grid Arrays (LGAs) assembled only with SAC305 solder paste than for the corresponding SAC305 based BGA assemblies. This superior performance was shown to be a direct effect of the solder microstructure. The final Sn solidification temperature strongly affects the initial microstructure of a SnAgCu solder joint, including the Sn grain morphology, and thus the thermomechanical behavior of the joint. Right after reflow, larger BGA joints of SnAgCu alloys, which solidify at higher temperature, reveal either a single β-Sn grain or three large grains with clearly defined boundaries formed by cyclic twinning. The orientations of the highly anisotropic Sn grains are not yet controllable in manufacturing, leading to substantial statistical scatter in the performance of the solder joints. Typical LGA solder joint dimensions, however, tend to facilitate greater undercooling and the formation of an alternative interlaced twinning microstructure. A systematic study was undertaken to identify the parameters that control the interlaced twinning microstructure. Sn grain structures were characterized by crossed polarizer microscopy and electron backscatter diffraction (EBSD). Precipitate sizes and distributions were measured using backscattered scanning electron microscopy and quantified using image analysis software. Systematic effects of solder alloy, dimensions and pad finishes were identified. Recommendations are made as to design and materials selection. The practicality of controlling the desired microstructure, as well as potential disadvantages for certain applications is discussed.
Keywords: BGA: Ball Grid Array; LGA: Land Grid Arrays; SAC305: Sn3Ag0.5Cu (wt %); EBSD: Electron Backscattered Diffraction ENIG: Electroless Nickel Immersion Gold; ENLG: Electrolytic Nickel Gold
Author: Martin Anselm
Abstract: In electronics the initial intermetallic formed on electroless Ni substrates during component ball attach can be exposed to high concentrations of Cu when components are assembled to Cu based circuit boards such as Immersion Ag, Immersion Sn, Cu OSP, or HASL. Ni solubility limits in SnPb37, SnAg3.5, and SAC304 are compared and determined to affect the kinetic intermetallic formation at the electroless Ni boundary.
Short and long reflow profiles are used to empirically control the Ni concentration within the alloys. All combinations were then subjected to limitless Cu in a second reflow. SEM/EDS, XPS, and EPMA were used to assess the initial intermetallic of the ball attach as compared to the intermetallic formed following the addition of limitless Cu. Addition of limitless Cu was introduced to the solder joint by either adding Cu powder or by using Cu pins in Hot Bump Pull (HBP) testing. Cold Bump Pull (CBP) and HBP testing was performed in order to characterize the mechanical behavior of the resulting intermetallic morphologies. Morphologies were then correlated to the mechanical failure mode and peak load to failure.
Ni solubility limits affect the chemical potential difference between initial ball attach and after the ball is allowed to become saturated with Ni in a long ball attach profile. In cases of higher Ni solubility (lead free) a dual phase intermetallic was observed, however in cases of low Ni solubility (SnPb37) a dual phase Ni-Sn intermetallic condition was not observed. It was observed that in SnPb little to no intermetallic morphology variations were produced with long and short ball attach profiles. In lead free however the higher solubility seemed to result in dramatic morphological differences during short and long reflow ball attach. The differences in morphology observed during ball attach continued following the addition of limitless Cu in a second reflow. These morphologies also resulted in marked differences in mechanical behavior ball pull testing. The results of this study present a favored material and reflow process for improved mechanical behavior.
Author: Linlin Yang and Peter Borgesen
Abstract: The present effort addresses effects of varying the cycling amplitude on the shear fatigue life of individual SnAgCu solder joints. So far studies of effects of combinations of amplitudes and loading types on the reliability of lead-free solder are extremely limited, and little work has been reported regarding extrapolations of test results to life in service. Various alloy compositions, geometries, surface finishes and pad sizes were tested to assess the validity of the linear damage accumulation rule almost universally applied in the industry. Variations in amplitude were seen to affect solder properties. A dry-section technique was developed for in-situ monitoring of microstructure evolution during cycling. Slip band evolution at various loading levels offered insight into potential reasons for the solder property changes. Preconditioning under a harsh load was found to change the acceleration factor in subsequent cycling.
Author: Linlin Yang and Peter Borgesen
Abstract: A major ongoing effort addresses the breakdown of Miner’s rule in the assessment of the fatigue life of lead free solder joints. This was originally observed for the life of BGA assemblies limited by the failure of the intermetallic bond to the pads, or by pad cratering, as well as for the fatigue life of the solder itself. In general, it appeared that Miner’s rule would underestimate life in a step-up test (low amplitude loading followed by cycling to failure with a higher amplitude) and overestimate it in a step-down test. As far as solder fatigue is concerned, the same trend was found in systematic joint level experiments. The opposite trend is, however, observed in single sided cycling bending of BGA assemblies. Indications are that this may be an artifact of the experimental conditions rather than a contradiction of previous results. The present effort also identifies the formation of slip bands in the SAC305 solder joints in very harsh cycling.
Author: Hohyong Lee and Seungbae Park
Abstract: Creep behavior of the solder joint is one of the critical concerns in reliability of the electronic package, especially above half of the homologous temperatures. Considering melting point of the lead free solder alloys, creep deformation can be dominantly occur even at room temperature.
Among the three phase of the creep curve, the steady state strain rate in secondary phase is the most important design parameter to predict life time of the solder joint. To measure the strain rate of the solder during the creep deformation, several studies conducted creep test using bulk solder samples. However, mechanical properties of the bulk solder are different from actual solder joint due to its microstructure and intermetallic layer. Thus, creep behavior of the actual solder ball need to be measured to obtain the realistic creep data. In this study, the constant creep rate of the actual solder ball in ball grid array (BGA) packages is measured using 2D Digital Image Correlation (DIC) measurement technique.
Three lead free BGA packages, SAC105, SAC305 and Sn/Cu/Ni alloy (SCN) are used for the creep test. While 9.78 N, 19.56 N, and 29.34 N of the compressive loads are applied at 20°C, 40°C, 60°C and 90°C to each type of the specimen, a microscopic images of the cross sectioned surface of the solder ball are captured. The series of images are sequentially taken during the predetermined time interval and transferred to DIC software to generate full-filled surface displacement data. From the displacement data, constant creep rate of the actual solder ball is acquired.
Author: Y. Jaradat, J. E. Owens, A. Qasaimeh, B. Arfaei, L. Yin, M. Anselm and P. Borgesen
Abstract: The fatigue and failure of interconnects under realistic service conditions are almost never well approximated by cycling with fixed amplitude, or in fact by a single loading condition. However, results of tests with varying amplitudes, not to mention tests combining vibration and thermal cycling, are not easily generalized.
We and others have reported on the breakdown of common damage accumulation rules in cycling of SnAgCu solder joints. A general trend is emerging for isothermal cycling where the remaining life, after a step-down in amplitude, is consistently shorter than predicted by Miner’s rule, while a step-up tends to have the opposite effect. These effects continue to increase with repeated alternations in amplitude. We show this to be consistent with the measured evolution in the associated solder properties. Similar trends were found for annealed Cu. We propose a practical approach to the assessment of life under any combination of amplitudes.
Keywords: Reliability, damage accumulation rules, service conditions, life in service, Pb-free.
Author: Linlin Yang, Awni Qasaimeh and Peter Borgesen
Abstract: Previous work has shown the life of a SnAgCu solder joint in a typical BGA or CSP assembly in thermal cycling to scale with the time to completion of a network of high angle grain boundaries across the high strain region of the joint. The associated recrystallization does, however, require significant temperature variations. Isothermal cycling leads to failure by transgranular crack growth which is certain to have very different acceleration factors. Accelerated thermal cycling tests are therefore only relevant to service conditions for which temperature excursions are still large enough to cause the formation of a recrystallized region across the joint. We present results indicating that cycling between 20C and 60C should be sufficient. Many practical applications involve a combination of thermal excursions and mechanical cycling, and there is little doubt that thermal cycling induced recrystallization will tend to lead to much faster crack growth through the solder in subsequent vibration, etc. We discuss how this greatly complicates the definition of a conservative but still practical accelerated test protocol.
Author: Pericles Kondos, Michael Meliunas and Martin Anselm
Abstract: Epoxy fluxes act like conventional fluxes in the initial stages of reflow, but then harden and form a “shell” around individual solder joints. This shell is expected to protect the joints from mechanical stresses, a protection not offered by conventional fluxes. For this reason, epoxy fluxes have been of increased interest lately in applications when use of capillary underfill after reflow is difficult or impossible to implement. Process development for the use of epoxy fluxes is not trivial and advantages in throughput and cost are most dramatic in dipping the material as compared to dispensing. However, reaching manufacturer’s recommended amounts of flux under the device can be challenging in dipping applications.
The assembly process was investigated during the initial stage of this project, and dipping and placement parameters that allowed consistent fluxing were determined. For this early research project only a single material was tested. Two different epoxy flux processing thicknesses were developed for a CSP and a WLCSP of slightly different sizes but the same pitch and ball size. The CSPs were placed on drop testing boards and were subjected to drop or bend tests.
Assemblies with the WLCSPs were subjected to accelerated thermal cycling. The reliability performance of all these parts was compared to conventionally-fluxed, nonunderfilled assemblies. Failure modes were studied and compared as well. It was seen that epoxy flux could indeed improve the performance as compared to non-underfilled parts if properly applied, but if not it might even reduce reliability.
Keywords: Epoxy flux, mechanical reliability, drop test, CSP, WLCSP, flux dipping, accelerated thermal cycling.
Author: Harry Scoeller and Michael Meilunas
Abstract: 0.3mm pitch re-ball assembly process and reliability is discussed. With 0.3mm pitch technology on the horizon for many OEMs rework is likely to be a major concern for high volume prodiuction. A manual rework process is developed using a standard assembly procedure. The drop test reliability of the reworked components is compared to its virgin counterpart. Reworked components had a slightly greater N63.2 life than virgin components.
Author: S. Joshi, B. Arfaei, A. Singh, M. Gharaibeh, M. Obaidat, A. Alazzam and M. Meilunas
Abstract: Land Grid Array packages are attractive for a number of reasons, including the small overall outline of assemblies and the ability of the user to choose the solder alloy. In general, life in thermal cycling may also not be reduced by as much as might be expected based on the standoff alone. In fact, we have shown packages soldered with SAC305 to survive as long as, or sometimes longer than, corresponding Ball Grid Array components in accelerated thermal cycling. This was found to be associated with a superior microstructure formed in the smaller solder volumes. This is, however, sensitive to materials and design parameters. So far we have been able to control it in our work. A major separate effort is focused on how to do so under general manufacturing relevant conditions. Even assuming the solder microstructure can be controlled questions remain as to whether it is also likely to be superior under realistic service conditions, whether it remains stable over time, and whether it has disadvantages for alternative damage and failure mechanisms such as solder pad cratering. This is the focus of the present report.
Keywords: Pb-Free Solders, Interlaced Twinning, Land Grid Array (LGA), Ball Grid Array (BGA), Pad Cratering.
Author: Michael Meilunas
Abstract: Printed circuit board assemblies can bend, or bow when attached to frames or slid into slots. The stress and strain created by deforming the boards is believed to be detrimental to component reliability. The experiment discussed in this report was designed to evaluate the effects of board bow between 0 and 2% using thermal cycle reliability data acquired on seven surface mount package designs.
This report has several associated documents including board datasheets and gerber files of the test board.
Author: Michael Meilunas
Abstract: The following report describes the experimental procedure and test results for a four-point cyclic bend evaluation used to compare the pad cratering resistance of six printed circuit board laminate materials.
Author: Michael Meilunas and Denis Barbini
Abstract: 0.3mm pitch Wafer Level and Chip Scale Package assemblies utilizing printing and dipping processes were evaluated for high-yield, high-throughput lead-free applications. Multiple stencil thickness, aperture sizes, paste/flux types, printer and placement machine settings were evaluated and the optimum, or near optimum, parameters were used to assemble components to multilayer test boards. Package designs evaluated included a 256 I/O standard array CSP, a 400+ I/O staggered array CSP and a 100 I/O WLCSP. Motherboards with open and copper filled microvia-in-pad were included in the comparisons. The completed test vehicles were subjected to mechanical shock (i.e. drop) testing, cyclic four-point bending, and/or thermal cycling in order to perform comparative reliability analyses.
Keywords: 0.3mm Pitch, Chip Scale Package (CSP), Wafer Level Chip Scale Package (WLCSP), SAC305, Accelerated Thermal Cycle (ATC), Drop Testing, Bend Testing.
Author: A. Singh and Michael Meilunas
Abstract: A cyclic 4-point bend test method was utilized as a means for investigating printed circuit board pad cratering. The maximum printed circuit board global strain and strain rates were varied to determine the effects of the parameters on failure time and mode. The test parameters were determined based on the allowable limits provided by the IPC/JEDEC 9702 and 9704 standards. Additionally, the effect of a single “prestress” bend intended to induce latent, or undetectable damage to the system was also evaluated using cyclic bend testing.
Author: Michael Meilunas
Abstract: A lead-free 0.45mm pitch photosensitive wafer-level chip scale package was provided to the AREA Consortium by a participating member for an assembly and reliability assessment. The device was quite thin, with a nominal silicon die thickness of just 45 microns and typical solder bump height of 230 microns. Variables evaluated included motherboard pad diameter, solder paste particle size, stencil aperture size and stencil thickness.
Author: Michael Meilunas
Abstract: Fine pitch surface mount devices often require printed circuit boards containing microvia-in-pad for complete route-out. Often the microvias are left open, or “unfilled”. The resulting hole in the printed circuit board pad can presumably alter the reliability of the assembly when compared to a plated shut, or “filled” via. This experiment, proposed and supported by an AREA member, was designed to compare the reliability of fine pitch (i.e. 0.5mm and below) assemblies created on filled and unfilled microvia-in pad using thermal cycling and drop testing.
Author: Martin Anselm and Wayne Jones
Abstract: As complexity in advanced manufacturing increases, especially for consumer electronics, the need to characterize the materials and processes used in electronic assembly also increases. OEM and EMS companies look to perform characterizations as early as possible in the process to be able to limit quality related issues and improve both assembly yields and ultimate device reliability. Many analytical methods are available to us on the market that each has their own risks and benefits. This paper will help identify some of these key limitations in the methods used for characterizing and evaluating solders, circuit board materials and surface finishes available in the market today.