microstructure and damage evolution in sn-ag-cu solder joints€¦ · and ag concentration. an...

0-7803-8906-9/05/$20.00 ©2005 IEEE 2005 Electronic Components and Technology Conference

Microstructure and Damage Evolution in Sn-Ag-Cu Solder Joints

L. P. Lehman, R. K. Kinyanjui, J. Wang, Y. Xing, L. Zavalij, P. Borgesen*, E. J. Cotts Department of Physics and the Materials Science Program

Binghamton University, SUNY 13902 *Universal Instruments Corporation; Binghamton, NY

Abstract

The implementation of no-Pb soldering is progressing rapidly, often without immediately obvious problems. However, our current understanding of pertinent materials issues is far from sufficient to prevent surprises, and potentially serious miscalculations. The fundamentally different nature of no-Pb from SnPb solder joints has serious consequences for the design and interpretation of accelerated tests.

The present work addresses the behavior of ball grid array (BGA), wafer level chip scale package (WL-CSP) and flip chip assemblies with SnAgCu solder joints in long term thermal cycling. Components were assembled onto either Cu or Ni pads on high-Tg printed circuit boards and cycled between 0oC and 100oC. Samples were removed at various stages for cross sectioning and microstructural characterization at various stages. Careful examination of the formation of intermetallic compounds at the pad surfaces, the size and orientation of the Sn grains, the morphology and number of precipitates, and the growth of cracks, revealed the sensitivity of the damage evolution to the solder microstructure. The consequences of the observed pattern of damage evolution are discussed.

Introduction After years of research and discussion, the transition of

the microelectronics industry to lead free soldering has become inevitable. The remaining question is one of timing, particularly for certain classes of products. Even military and implantable medical electronics suppliers, (many of whom could be granted legal exemptions), will be facing a gradual loss of manufacturing infrastructure if they do not switch to lead free components. While a growing number of companies are already making the transition, it is becoming increasingly clear that only a limited knowledge and data base has been established for the implementation of lead free soldering. A range of potentially critical surprises may be in the offing.

There are indications that structures may be weakened in lead free soldering in ways specific to the combination of component, substrate, and process. Such degradation is often not detectable in standard tests, but leads to an occasional loss of performance at later times. Most recently, observations of occasional degradation and brittle failure of the intermetallic structures formed by soldering has caused disquiet across the industry, especially because none of the currently realistic solder pad finishes seem to be consistently immune to this problem [1-4].

The properties and behavior of actual, lead free solder joints is the focus of the most concern. The microstructures of lead free solder joints are observed to significantly vary within a single assembly, undoubtedly affecting the

corresponding deformation properties. As the present work reveals, variations in the microstructure of lead free solder joints affect the damage evolution paths in cycling. One might expect that the combination of this variability with the effect of the stress distribution across an assembly, would result in considerably broader failure distributions than those observed for PbSn components. Somewhat surprisingly, broader failure distribution are not often observed when testing a finite number of samples. But, as seen in the present work, the behavior of a few joints may be that of clear statistical 'outliers'. In such joints it is observed that damage progresses much faster than in neighboring ones. Combined with the statistics of location for the much larger sample size representative of actual manufacturing, this would provide for occasional, very early failures that may not be captured by extrapolation of accelerated test results. (This should not be confused with the question of extrapolating to 'service' conditions).

The performance of SnPb based solder joints is often taken as a critical reference point for reliability. Unfortunately, simple comparisons of SnPb and Pb free are often misleading. Realistic no-Pb solder joints are proving sensitive to parameters that were not important for SnPb joints. Standard accelerated tests often favor no-Pb joints over SnPb ones, even when the latter may do better in service. Current proprietary research is showing all of this to be associated with the sensitivity of no-Pb solder joint properties to microstructure [5].

The result of most importance to the industry is the performance of lead free solder joints under service conditions. Thus accelerated test results are meaningless without at least an implicit extrapolation to such service conditions. In the case of SnPb solder joints, consensus has been attained as to the required accelerated test performance for a given type of application. Essentially, stress/strain analysis and modeling has allowed the scaling of over thirty years of SnPb solder fatigue data for a wide variety of joints onto a Master curve, and thus the extraction of semi-empirical damage functions and 'acceleration factors'. In addition, experiences with the life of various SnPb products in actual service have offered opportunities to check predicted values. For SnPb solder fatigue, similar mechanical treatment usually leads to similar fatigue crack growth rates, without many surprises.

Even if a data base similar to that for SnPb was available for no-Pb solder joints, a scaling onto a Master curve would be considerably more challenging. Recent research has shown lead free solder joint microstructures to vary quite dramatically with joint size, because of both intermetallic formation at the pad surfaces, and a volume dependent degree

674

2005 Electronic Components and Technology Conference

of undercooling after reflow. The microstructure is also significantly affected by temperature and stress enhanced diffusion during thermal cycling. It is apparent that considerable mechanistic insight will be required to establish meaningful acceleration factors or a damage function for lead free solders.

Below, we present results of a study on the evolution of damage in SnAgCu solder joints during a particular 'standard' thermal cycling test. BGA, WL-CSP and flip chips were attached to Ni and Cu pads on high-Tg FR-4 boards with SnAgCu solder joints and cycled between 0oC and 100oC. The 20 minute cycle had 5 minute ramps and 5 minute holds at the two temperatures. Assemblies were removed for cross sectioning and microstructural characterization at various stages of cycling. Damage accumulation was examined using optical and electron microscopy techniques. These tools allowed a careful examination of 1. formation of intermetallic compounds at

solder/metallization interfaces, 2. the size and orientation of Sn grains in the SnAgCu solder, 3. the morphology and number of intermetallic precipitates in

the SnAgCu solder matrix, and 4. crack propagation in the solder joint.

Background The relatively large number of microstructural factors

which can vary in SnAgCu is in contrast to the case of PbSn solders, which contain no equilibrium intermetallic compounds, and which typically display less dramatic variation in grain sizes. Thus it is important to carefully consider the evolution of the microstructure of SnAgCu alloys when considering their reliability [6-18].

The Cu in SnAgCu solder can diffuse rapidly and react with Ni metallizations, and the Sn in the solder, to form (CuNi)6Sn5 at solder/Ni interfaces, thus decreasing the Cu content of the SnAgCu solder[6-10]. As the mechanical properties of a soft material such as Sn can depend upon the number and size of precipitates of greater hardness such as Cu6Sn5, changing the Cu content can affect the mechanical response of SnAgCu solder. The thickness, adhesion, morphology and composition of the intermetallic compounds at SnAgCu solder/metallization interfaces may also affect the mechanical reliability of the solder joint.

The number and orientation of the Sn grains in a SnAgCu solder joint are also potentially very important to the mechanical response of a SnAgCu solder joint. Under some conditions, a SnAgCu solder joint may form with a small number of grains, or only one grain. As the mechanical properties of Sn are highly anisotropic,[13] devices with Sn as a main constituent which display relatively large Sn grains may exhibit a wide range of mechanical behaviors, each corresponding to a particular orientation of a large Sn grain. The Sn grain size has been shown to be very sensitive to Cu and Ag concentration. An investigation of the microstructure of Sn-Ag-Cu solder using scanning electron microscopy, cross polarized light optical microscopy and pole figures from X-ray diffraction, for compositions near the ternary eutectic (approximately: Sn-3.5Ag-0.9Cu-wt.% or Sn-3.8Ag-1.7Cu at.%) found order of magnitude changes in Sn grain size with

small (one or two percent) changes in the Cu concentration[8]. For instance, the optical micrographs of Fig. 1 for a series of Sn-Ag-Cu samples of constant Ag concentration (3.9 weight percent) and varying Cu concentration (0 weight percentage in Fig. 1(a), 0.27 weight percentage in Fig. 1(b), 0.59 weight percentage in Fig. 1(c), and 1.1 weight percentage in Fig. 1(d)) reveals the distinct dependence of Sn grain size on composition.

The size and number of equilibrium precipitates (Ag3Sn

and Cu6Sn5 in the Sn matrix of the SnAgCu solder) can also affect the mechanical response of the solder[13]. In the SnAgCu near eutectic system, two distinctly different size ranges of precipitates have been observed, those on order of one mm in spatial extent, and those on order of one to ten microns[8]. The details of the solidification process upon cooling from the melt are most important in determining this distinction in these near eutectic SnAgCu alloys. Upon cooling below the equilibrium liquidus temperature, Ag3Sn and/or Cu6Sn5 nucleate relatively easily. In contrast, the remaining Sn solution, Sn(Cu,Ag), displays marked undercooling.[8,9] The intermetallic compounds which nucleate in liquid Sn(Cu,Ag) typically have minutes to grow before Sn solidification slows this growth process. The solidification of Sn results in the formation a dendritic Sn structure, with the dendrites decorated with the smaller scale Cu6Sn5 and Ag3Sn precipitates. Thus a sample often has a wide range of sizes of precipitates, as seen in Fig. 2.

The SEM Back Scatter electron image in Fig. 2 provides some examples of the distinct varieties in the length scales of the Cu6Sn5 and Ag3Sn precipitates found in these SnAgCu near eutectic microstructures. The micrograph in Fig. 2 represents a single region of a sample which had an average composition of Sn-3.0Ag-1.1Cu by weight percent, and was cooled from the melt at a rate of 0.1 oC/s, after being heated to a temperature of 250 oC. The micrograph reveals Sn dendrites

Figure 1. Laboratory produced samples of SnAgCu solder. Optical polarized light micrographs provide contrast due to grain orientation; a) Sn-3.0Ag, b) Sn-3.0Ag- 0.3Cu, c) Sn-3.0Ag- 0.6Cu, d) Sn-3.0Ag- 1.1Cu (see ref. [8]).

675


ringed by intermetallic precipitates with a length scale of approximately 10 microns. Very large Ag3Sn or Cu6Sn5 precipitates, on order of 1 mm, are also observed. The large aspect ratio Ag3Sn precipitate has a shape consistent with the Ag3Sn plates previously reported, while the shapes of Cu6Sn5 precipitates are consistent with the rod like geometries reported a number of times in the literature. As is apparent from this micrograph, such precipitates can be quite large.

The size of the Sn dendrites in SnAgCu solder can also

affect the mechanical properties of these alloys[9]. Sn dendrite arm size have been reported[9] to distinctly decrease with decreasing SnAgCu solder ball size, i.e. flip chip samples have distinctly smaller dendrite sizes than ball grid array samples, for similar cooling rates, as seen in Fig. 3.

While these differences were distinct, it was also noted that relatively large distributions in the undercooling temperatures were observed for the same ball size and cooling rate. That is to say, some distinct variations in microstructure are to be expected for similar SnAgCu samples treated under similar annealing conditions. This paper considers how variations of composition, size and thermal history of Sn-Ag-Cu solders affects solder joint microstructure, and subsequent damage accumulation

Experimental Sn-Ag-Cu solder materials were derived from two main

sources, commercially available solder joints and solder ingots fabricated in the laboratory. Laboratory prepared solder ingots, approximately one gram each, were carefully weighed from pure elements that were melted together in an oxygen gettered, argon atmosphere using an arc-melting furnace. These ingots were then turned over and remelted in the arc-melter. After characterization, fractions of an ingot were cooled at controlled rates near 1oC/s to produce samples. The commercially available solder joints included ball grid array (BGA), chip scale packaging (CSP), and flip chip components in SnAgCu solder joints. All components were assembled on either OSP-coated copper, or electroless-nickel-immersion-gold (ENIG), coated pads on high-Tg FR-4 boards. The commercial solder joint samples were subjected to air-to-air thermal cycling between 0oC and 100oC. The 20 minute cycle had 5 minute ramps and 5 minute holds at the two temperatures. Assemblies were removed for cross sectioning and microstructural characterization at various stages of cycling.

Samples were examined with an optical microscope under bright field and cross polarized light imaging in order to identify and locate the different phases and Sn grains. This is possible since Sn is a birefringent material, so polarized light experiences a small rotation of the polarization angle on reflection from Sn crystals. By viewing the reflected light through an analyzing polarizer, crossed with-respect-to the incident polarized light, contrast can be seen that depends on the grain orientation.[8,9]

An Electron Microprobe Analyzer was used to determine the chemical composition of phases in some cases. In other cases, scanning electron microscopy (SEM) using backscatter composition mode imaging and energy dispersive analysis (EDS) were used to identify the composition of phases and structures not immediately identifiable from their morphologies.

Results and Discussion BGA, WL-CSP and flip chip assemblies were chosen to

represent very different solder joint sizes. Most of the flip chips were cycled without underfilling to allow direct comparisons of the results to those for the other components. Cycling to failure of 16 samples of each, a quite typical sample size in accelerated testing, led to reasonably narrow Weibull distributions with characteristic life values of 15,000 cycles, 516 cycles and 55 cycles. The simple construction of the WL-CSP components allow for particularly simple comparisons with the flip chip results, and the results for both were found to fall on a simple empirical scaling with those for

Figure 2. Example of different length scales of intermetallic precipitates in SnAgCu solder samples. SEM Back Scatter composition mode image provides contrast due to average atomic number. Sn-3.0Ag-1.1Cu shows large primary precipitates and Sn dendrites surrounded by secondary precipitates.

Figure 3. The effect of SnAgCu sample size on microstructure. Three SEM back scatter composition mode images showing microstructure of three commercial made lead-free solder joints, a) BGA(ball diameter 760 micron), b) WL/CSP(ball diameter 265 micron),, c) Flip-Chip (ball diameter 125 micron),. All are un-aged with no ATC cycles. (see ref. [9])

676


similar components in terms of solder joint size and distance to neutral point. As discussed above this may in fact effectively include an effect of size dependent microstructure, but we shall be more concerned with statistical variations and outliers.

As Fabricated Samples –No Thermal Cycling A number of SnAgCu solder balls in commercial solder

joints were examined. The components in these solder joints generally had Ni(V) metallizations, while the substrate metallizations were either Ni(Au) or Cu OSP. As is evident in the optical micrographs using crossed polarizers of Figs. 4-7, the distinct microstructures observed previously in laboratory samples of SnAgCu alloys of various compositions (cf. Fig. 1) are observed in commercial solder balls as well. In particular, large variations in the Sn grain sizes are observed. Fig. 4 is a optical micrograph with cross polarizers of a BGA sample. It is clear that there are only a small number of Sn grains in this large solder ball. Furthermore, the variation of the apparent color of the Sn grains in the crossed polarized image, and the angles between the grains, both correspond very well to those observed in cyclic twins; i.e. these Sn grains are highly oriented. Simlar observations are sometimes found for flip chip and CSP samples (Figs. 5-7), where once again Sn grains which are large relative to the sample size are observed. In these smaller CSP and flip chip samples an “interlaced” Sn grain microstructure occurs more frequently, as is seen in Figs. 5 and 6 near the substrate and/or the component side. Fig. 7 reveals the grain structures for some flip chip samples.

The occurrence of an interlaced microstructure in smaller SnAgCu samples (CSP and FC) more frequently than larger SnAgCu balls (BGA) is consistent with the correlation[8] of the interlaced microstructure with less Cu rich Sn (e.g. SnAg eutectic compositions) as illustrated in Fig. 1. The observed dependence[9] of degree of undercooling of a SnAgCu solder ball with inverse nominal diameter means that the smaller CSP and FC balls will undercool much more than the BGA balls, in general, as previously reported. The greater degree of

Figure 4. Optical polarized light image of Sn-3.0Ag-0.8Cu commercial BGA joint. Contrast from grain orientation. The sample was not aged, nor was it cycled.

Figure 5. Optical polarized light image of Sn-3.5Ag-1.0Cu commercial WL/CSP joint. Contrast from grain orientation. Un-aged and no ATC cycles.

Figure 6. Optical polarized light image of Sn-3.5Ag-1.0Cu commercial WL/CSP joint. Contrast from grain orientation. Un-aged and no ATC cycles.

Figure 7. Optical polarized light image of Sn-3.5Ag-1.0Cu commercial Flip-Chip joints. Contrast from grain orientation. (a)Aged 50 hrs at 150ºC and cycled 10 ATC cycles, 0-100ºC. (b) Un-aged and no ATC cycles.

677


undercooling means that the liquid Sn(Cu,Ag) in these smaller SnAgCu solder balls will be much less rich in Ag and Cu upon solidification. This can be understood with consideration of the metastable liquidus surfaces for this ternary system; consistent with the binary Ag-Sn and Cu-Sn systems[8,9] .

Effects of Thermal Cycling In most cases the failure mechanism, as manifested in

terms of crack initiation and progression, displayed many similarities with the canonical behavior invariably observed for PbSn solder joints. However, even within the relatively small number of samples examined in this study (less than one thousand joints) several CSP and flip chip solder joints were found to exhibit a substantially faster damage evolution than other joints at similar times and locations. This did not necessarily mean that the overall assembly would fail faster, as the thermal cycling induced loads on the individual joints usually vary strongly with location across an area array assembly. This is particularly true for components such as the present WL-CSPs and (non-underfilled) flip chips where the loads increase strongly with the 'distance-to-neutral-point' (DNP). In fact the mere observation of an increased variability, compared to assemblies soldered with PbSn, in the location of the first joint to fail, albeit in most cases still within a region quite near the greatest DNP, is a clear indication of a significant effect of joint microstructure. Anyway, in the rather rare cases where a particularly 'weak' microstructure coincides with the greatest DNP we may expect very early failure indeed. In the following we therefore pay particular attention to such microstructures.

Upon examining a number of SnAgCu samples, it is clear that crack initiation generally occurs at the component side. Cracks generally form near the edge of the intermetallic compound/solder interface, though sometimes further in to the solder. Often these cracks propagate in a bow shape, away from the interface, such that a complete crack often resembles that of Fig. 8. But the SnAgCu system exhibits a number of samples which reveal crack propagation paths which differ substantially from the identifiable, general trends. For instance, Fig. 9 reveals a crack path in an underfilled flip chip sample which transverses the entire joint.

In general (though once again, only in general) our observations to date upon thermally cycling SnAgCu solder joints indicate a systematic progression of the length of a crack. That is to say, crack length increases with the number of ATC cycles, and with position (i.e. with increasing DNP). The balls closest to the corner of the component generally are the first to show crack initiation. As the number of cycles increase, the number of balls observed to have cracks increases, generally in a systematic progression away from the corner of the component. Variations from this general progression are often correlated with the presence of grain boundaries, (and sometimes with the presence of voids) near the solder ball edge. For instance, in Fig. 10, a flip chip sample with a Cu OSP metallization on the substrate, it is clear that the crack does not meet the corner of the solder ball. Of course, this cross section is a two dimensional slice of the sample, and the crack may have begun at the corner of the

ball elsewhere on the joint, and propagated upwards to the point of observation in Fig. 10. It is interesting that a optical microscopy examination with crossed polarizers of this cross

Figure 8. Optical images of Sn-3.5Ag-1.0Cu commercial Flip-Chip joint, un-aged and 60 ATC cycles. (a)Bright field image shows microstructure. (b) polarized light image, contrast from grain orientation.

Figure 9. SEM Secondary Electron image of Flip chip joint reveals crack path which subtends the SnAgCu solder joint.

Figure 10. Sn-3.5Ag-1.0Cu commercial Flip-Chip joint, un-aged and 60 ATC cycles. (a)Bright field image shows microstructure. (b) polarized light image, contrast from grain orientation.

678


section (Fig. 10(b)) reveals that the crack is following a grain boundary. This is evidenced by the difference in coloration on either side of the crack.

The exceptions to a simple, systematic crack progression

are perhaps the most interesting examples of the behavior of this system. They are perhaps more prevalent than one might expect. For instance, in our sample of approximately 500 flip chip, lead free solder joints, 14 crack sooner than would be expected in this simple picture. For instance, Fig. 11 reveals a FC solder joint which revealed the initiation of a crack after ten thermal cycles in the ball position eight balls from the component corner. None of the other balls showed signs of cracks in this sample. For this ball position, cracks were not seen in other samples until three or four times as many cycles had occured. As can be seen in Fig. 11(b), the crack, in this cross section, is progressing through a large single grain. There is evidence that the crack began near the corner of the ball. The crack may be proceeding parallel to the direction of growth of from the melt of the Sn grain. This example is illustrative of the complex nature and influence of the microstructure of these solder joints.

Exceptions to the general progression of failures also

The exceptions to a simple, systematic crack progression include smaller magnitude effects. For instance, balls internal to the component cracking before their counterparts closer to, or at, the component corners. Fig. 12 reveals a crack at the edge of a CSP solder joint. This is notable in that a number of solder joints closer to the corner of the component in this assembly did not show any signs of cracks after the same number of cycles. The cross polarizer image for this sample

(Fig. 12 (b)) reveals a large number of orientations of Sn grains in the ball, in contrast to the case of the CSP ball closest to the corner for this sample (Fig. 13), which cross polarizers indicate has fewer Sn grains, in particular near the component (a region which is apparently single grained).

Figure 14. Sn-3.5Ag-1.0Cu commercial WL/CSP joint, un-aged and 250 ATC cycles. (a)Bright field image shows microstructure. (b) Polarized light image, contrast from grain orientation.

Another example of the influence of grain boundaries on the response to stress, and on crack propagation is presented in Fig. 15, for a CSP solder joint. In this figure the crossed polarizers image shows that one grain boundary subtends the joint. It is evident that the solder ball has been sliding on this grain boundary, as there is evident displacement at the intersections of the grain boundary and the edge of the ball. Never-the-less, the most evident crack in this sample (see bright field image, Fig. 15(a)) is near the component!

Figure 11. Sn-3.5Ag-1.0Cu commercial Flip-Chip joint, aged 50 hours and 60 ATC cycles. (a)Bright field image shows microstructure. (b) polarized light image, contrast from grain orientation.

Figure 12. Sn-3.5Ag-1.0Cu commercial WL/CSP joint, un-aged and 100 ATC cycles. (a)Bright field image shows microstructure. (b) polarized light image, contrast from grain orientation.

Figure 13. Sn-3.5Ag-1.0Cu commercial WL/CSP joint, un-aged and 100 ATC cycles. Polarized light image, contrast from grain orientation.

679


Examination of PBGA components that had been cycled up

to 2500 times revealed no cracking, nor incipient cracking. As noted in the text and reflected in Fig. 1, larger grain structures in this system are more generally found in samples with higher Cu and/or Ag content in the Sn melt before solidification. As discussed above, larger samples (i.e. PBGA) undercool less, and thus have higher Cu and Ag content in the Sn melt before solidification than CSP and FC samples. In fact, larger grains are observed in PBGA samples (cf. Fig. 4). Further work is in progress in an effort to understand any possible relation between grain size and time to failure in PBGA samples.

Conclusions As expected, the solder microstructure does have a

significant effect on the propagation of cracks in thermal cycling of SnAgCu solder joints. This is reflected in a scatter in the location of the first solder joint to fail which goes beyond the statistical variations typically observed for SnPb. Still, the formation and propagation of cracks usually does tend to proceed in a rather systematic fashion, increasing with time and distance to the neutral point (location of higher loading). In fact, in spite of the scatter overlaid on this by the variations in microstructure, typical failure distributions remain surprisingly narrow.

This is not necessarily good news. The occasional observation of a clear outlier, i.e. a joint exhibiting much greater damage than neighboring ones or joints in higher-stress locations, shows that particularly early failures must occur in the relatively rare occasions when a similar solder joint microstructure happens to coincide with the location of highest load. Such early failures may be unlikely to be captured within a typical accelerated test sample size, or even be predictable based on extrapolation along a Weibull distribution. However, the observation of such particularly 'weak' joint structures within a total sample of less than a thousand joints shows that the corresponding early failures will likely occur repeatedly within a typical manufacturing volume. An occasional strong effect associated with solder joint microstructure is further disconcerting because of the established variation in microstructure with solder joint size, as this makes comparisons between different assemblies much more difficult and thus further limits the statistical significance of our data base.

Acknowledgments The authors are pleased to acknowledge the financial

support for the research by the National Science Foundation (DMI 0218129).

References 1. Borgesen, P., and Henderson, D. W., "Fragility of Pb-free

solder joints", white paper, Aug. 24, 2004, www.uic.com 2. Chiu, C., Zeng, K., Stierman, R. Edwards, D., and Ano,

K., “Effect of Thermal Aging on Board Level Drop Reliability for Pb-free BGA Packages”, 54th ECTC Conf., 2004

3. Date, M., Shoji, T., Fujiyoshi, M., Sato, K., and Tu, K. N., “Impact Reliability of Solder Joints”, 54th ECTC Conf., 2004

4. Gregorich, T., Holmes. P., Lee, J., and Lee, C., “SnNi and SnNiCu Intermetallic Compounds Found When Using SnAgCu Solders”, IPC/Soldertec Global 2nd Int. Conf. on Lead Free Electronics, June 2004

5. Area Array Consortium research, Universal Instruments Corporation, Binghamton, NY

6. Zribi, L. Zavalij, P. Borgesen, and E. J. Cotts, “The Growth of Intermetallic Compounds at Sn-Ag-Cu Solder/Cu and Sn-Ag-Cu Solder/Ni Interfaces and the Associated Evolution of Microstructure,” Symposium Y, Proceedings of the Fall Meeting of the Materials Research Society, Y8.10, Vol. 652 (1999).

7. Cotts, E. J, Chromik, R, Borgesen, P, Kinyanjui, Rand Zribi, A; Formation of Intermetallic Compounds at Pb-Sn/Metal and Lead-Free/Metal Interfaces in Solder Joints”, Published in the Handbook of Lead-Free Solder Technology for Microelectronic Assemblies, Edited by Puttlitz, K.J and Stalter, K. A, 2004 by Marcel Dekker, Inc.

8. L. P. Lehman, S. N. Athavale, T. Z. Fullem, A.Giamis, R. Kinyanjui, M. Lowenstein, K. Mather, R. Patel, D. Rae, J. Wang, Y. Xing, L. Zavalij, P. Borgesen, E. J. Cotts, “Growth of Sn and Intermetallic Compounds in Sn-Ag-Cu Solder” J. Electronic Materials 33, 1429(2004).

9. R. Kinyanjui, L. P. Lehman, L. Zavalij, E. Cotts “Effect of Sample Size on the Solidification Temperature and Microstructure of SnAgCu near Eutectic Alloys”submitted to Journal of Materials Research.

10. L. P. Lehman, R. K. Kinyanjui, L. Zavalij, A. Zribi, E. J. Cotts, "Growth and Selection of Intermetallic Species in Sn-Ag-Cu No-Pb Solder Systems based on Pad Metallurgies and Thermal Histories", 2003 Electronic Components and Technology Conference, pp1215-1221, May 27-30, 2003, New Orleans LA.

11.Henderson, D. W, Woods, J J, Gosselin, T. A, Bartelo, J, King, D. E, Korhonen, T. M and M. A. Korhonen, M. A, L. P. Lehman, E. J. Cotts, Sung K. Kang, Paul A. Lauro, Da-Yuan Shih, Charles Goldsmith, and Karl J. Puttlitz; “The Microstructure of Sn in Near Eutectic Sn-Ag-Cu Alloy Solder Joints And Its Role In Thermomechanical Fatigue”, Journal of Materials Research 19,(2004).

12. Henderson, D. W, Gosselin, T, Sarkhel, A, Kang, S. K, Won-Kyoung- Choi, Da-Yuan-Shih, Goldsmith, C and

Figure 15. Sn-3.5Ag-1.0Cu commercial WL/CSP joint, un-aged and 150 ATC cycles. (a)Bright field image shows microstructure. (b) Polarized light image, contrast from grain orientation.

680


Puttlitz, K. J, “Ag3Sn plate Formation in the Solidification of Near ternary Eutectic Sn-Ag-Cu Alloys” Journal of Materials Research, 2775 (2002).

13 .Telang, A. U, Bieler, T.R, Choi, S and Subramanian, K.N; “Orientation Imaging Studies of Sn-Based Electronic Solder Joints”, Journal of Materials Research 17, 2294 (2002).

14. Moon, Z. W. Boettinger, W. J, Kitten, U. R. Biancaniello, F. S and Handwerker, C. A, JEM, Vol. 29, No. 10, Pp.1122, 2000 and Loomans, M. E and Fine, M. E, Metall. Mater. Trans. A, Phys. Matall. Mater. Sci, Vol. 31A, No.4, Pp.1155, 2000.

15. S. Huh, K. Kim and K. Suganuma, “Effect of Ag Addition on the Microstructural and Mechanical Properties of Sn-Cu Eutectic Solder”, Materials Transactions 42, 739(2001).

16.. Yoshiharu Kariya, Yasunori Hirata and Masahisa Otsuka & "Mechanical Fatigue Characteristics of Sn3.5AgX (X=Bi, Cu, Zn and In) Solder Alloys, Journal of Electronic Materials 28, 123-1269 (1999)

17. Moore, K.I Zhang, D.L and Cantor, B; “Solidification of Pb Particles Embedded in Al”, Acta Metall. Mater. 38, 1327 (1990).

18. D. Turnbull, "Isothermal rate of solidification of small droplets of mercury and tin," Journal of Chemical Physics 18,768 (1950).

681

microstructure and damage evolution in sn-ag-cu solder joints€¦ · and ag concentration. an...

Documents