Thermomigration in lead-free solder joints. Mohd F. Abdulhamid, Shidong Li and Cemal Basaran*

Transcription

1 Int. J. Materials and Structural Integrity, Vol. 2, Nos. 1/2, Thermomigration in lead-free solder joints Mohd F. Abdulhamid, Shidong Li and Cemal Basaran* Electronic Packaging Laboratory University at Buffalo, SUNY 243 Ketter Hall, Bufallo, NY 14260, USA Fax: (716) *Corresponding author Abstract: In the next generation nanoelectronics and SiC based electronic packaging, current density and temperature gradient will be larger in orders of magnitude. Electromigration and thermomigration are considered to be major road blocks leading to realisation of nanoelectronics and SiC based high temperature power electronics. In this paper, damage mechanics of 95.5Sn4Ag0.5Cu (SAC405) lead-free solder joints under high temperature gradients have been studied. This paper presents observations on samples which were subjected to 1000 C/cm thermal gradient for two hours, 286 hours, 712 hours and 1156 hours. It was observed that samples subjected to thermal gradient did not develop a Cu 3 Sn intermetallic compound (IMC) layer at the hot side due to Cu migration to the cold side thus causing insufficient Cu mass concentration to form Cu 3 Sn. On the other hand, in samples subjected to isothermal annealing exhibited IMC growth. In samples subjected to thermomigration, near the cold side the Cu concentration is significantly higher, compared to hot side. Extensive surface hardness testing showed increase in hardness from the hot to cold sides, which indicates vacancy migration and Sn grain coarsening are in the opposing direction. Keywords: thermomigration; lead-free; hardness; grain coarsening; vacancy migration. Reference to this paper should be made as follows: Abdulhamid, M.F., Li, S. and Basaran, C. (2008) Thermomigration in lead-free solder joints, Int. J. Materials and Structural Integrity, Vol. 2, Nos. 1/2, pp Biographical notes: Mohd Abdulhamid is a Lecturer in the Faculty of Mechanical Engineering, Universiti Teknologi, Johor, Malaysia. He received his PhD from State University of New York at Buffalo. Shidong Li is a Research Assistant in the Electronic Packaging Laboratory at the State University of New York at Buffalo. Cemal Basaran is Professor and Director of Electronic Packaging Laboratory at the State University of New York at Buffalo, He specialises in experimental and computational damage mechanics of nano and power electronics packaging. He has authored more than 140 publications in the field of high sensitivity moiré interferometry inspection and Copyright 2008 Inderscience Enterprises Ltd.

2 12 M. Abdulhamid, S. Li and C. Basaran thermodynamics based damage mechanics of electronics packaging under electromigration, thermomigration and thermo-mechanical loads. He holds a MS degree in Civil Engineering from MIT and a PhD in Engineering Mechanics from the University of Arizona in Tucson. He is 1997 recipient of the Department of Defense ONR Young Investigator Award. He is the Assoc. Editor of ASME Journal of Electronic Packaging, IEEE Trans. on Advanced Packaging, and the Regional Editor for Americas for International Journal of Materials and Structural Integrity and a Member of Editorial Board of Open Civil Engineering Journal and Journal of Recent Patents on Electrical Engineering. 1 Introduction The insatiable demand for miniaturisation of electronics and high temperature power electronics is increasing electrical current density and thermal gradient in electronic packaging solder joints by orders of magnitude. It is well known that the high current density and high temperature gradient induce mass migration that degrades the solder joints (Ye, 2004; Ye et al., 2003a; 2003e). The two primary mass migration processes that manifest in solder joints subjected to high current density are electromigration and thermomigration (Ye, 2004; Ye et al., 2003a; 2003b; 2003c; 2003e; Basaran et al., 2003; 2005; Huang et al., 2006; ). When a metal conductor is subject to a high current density, because of scattering effect, the so-called valance electron-wind transfers part of the momentum to the atoms (or ions) of metal (or alloy) to make the atoms (or ions) move in the direction of the valance electrons. Electromigration causes mass accumulation and hillocks formation in the anode side, and vacancy condensation and void formation in the cathode side (Ye et al., 2003a; 2003e). Both hillocks and voids will cause the degradation of the solder joint and eventual failure. Thermomigration is a mass migration process that takes place due to high thermal gradients. Thermomigration takes place independent of electromigration however. Thermal gradient can be a result of Joule heating in the solder joint. Thermomigration is very similar to the Soret effect in fluids. Soret (1879) discovered that concentration of a salt solution in a tube with both ends at different temperatures does not remain uniform. The salt was less concentrated on the hot end compared to the cold end. He concluded that a flux of salt was generated by a temperature gradient, which results in a concentration gradient in steady state conditions (Platten, 2006). Ye et al. (2003e) showed that the same process also takes place in solder alloys. Experimental studies suggest that the effect of thermomigration on solder joint reliability is as serious as that of electromigration (Ye et al., 2003b; 2003c; 2003e; Huang et al., 2006; Roush and Jaspal, 1982). Thermomigration in PbIn solder was reported by Roush and Jaspal (1982) at a temperature gradient of 1200 C/cm. Both In and Pb move in the direction of the thermal gradient that is from hot to the cold. Huang et al. (2006) investigated thermomigration of SnPb solder alloy at an estimated temperature gradient of 1000 C/cm. They found that Sn moved to the hot end while Pb moved to cold end. Ye et al. (2003e) reported thermomigration may assist electromigration if the higher temperature side coincides with the cathode side, and counter electromigration if the hot side is the anode side.

3 Thermomigration in lead-free solder joints 13 In this paper, thermomigration effects on lead-free solder joints were studied experimentally. High thermal gradients, as high as 1000 C/cm, are applied to SnAgCu solder joints without any current flow. In other words, samples in our experiment were solely subjected to high temperature gradients but no electrical current in order to separate the influence of thermomigration from electromigration. All other studies published in the literature involve electromigration and thermomigration at the same time. The void nucleation, microstructural changes and intermetallic compound (IMC) growth on both sides of the solder joint were observed using scanning electron microscope (SEM) equipped with electron dispersive x-ray (EDX). The hardness properties are measured using nano indentation method. 2 Experimental setup The lead free solder alloy used in this study is commercial grade and has a composition by weight percentage of 95.5% Sn, 4.0% Ag and 0.5% Cu (SAC 405) with a melting point of 225 C. The solder ball joins two copper plates with dimensions of 19 mm by 38 mm and 0.8 mm thick. The copper plates are polished to mirror-like finish to remove any possible oxidation. The plate is then covered with solder mask except at locations where the solder joints are reflowed. Two one-mm thick glass plates are used as spacers to maintain the gap between the copper plates. The spacer also helps maintain a consistent solder joint height. The solder joints are reflowed using the JEDEC (2004) reflow profile. Four samples are sandwiched between a lower Al block in contact with a hot plate and an upper Al block in contact with a thermoelectric cooler as shown in Figure 1. Every contact surface is coated with thermal grease to maximise heat transfer. The open area between plates and between blocks is insulated to minimise effects from heat radiation and circulating heat flow. The hot side and cold side temperatures are maintained at 160 C and 50 C, respectively. Thermocouple probes are placed in holes, very close to test vehicle, on the Al blocks as shown in Figure 1. The temperatures are recorded every fifteen minutes. During this project, 24 samples were subjected to thermal stresses for duration ranges from two hours up to 1156 hours. 3 Sample preparation The thermally stressed samples were embedded in epoxy, sliced and mirror polished to the centre section of the joint using automated polishing machine with programmable polishing head. The final surface finish was polished using 0.05-micron silica suspension to ensure accurate nano indentation test results. Microstructural morphology and IMC elemental analyses were performed on the prepared samples using backscatter SEM equipped with EDX. Nano indentation tests were carried out using MTS Nano Indenter XP system with Berkovich tip at room temperature. Five rows of up to 25 indentations were used to evaluate the solder joint mechanical properties. Each row represents normalised distance of 0.1, 0.3, 0.5, 0.7 and 0.95 from the hot side (bottom side), as shown in Figure 2.

4 14 M. Abdulhamid, S. Li and C. Basaran Nanoindentation hardness measurements were performed using a maximum load of 250 mn, and loading time of 15s. The theory behind surface hardness measurement and elastic modulus calculations are presented in great detail in Oliver and Pharr (1992). Figure 1 (a) Front view of the test vehicle sandwiched between heating and cooling plates (b) Partial side view of test apparatus showing three out of four possible test vehicles (a) Front view (b) Partial side view Note: In both views, the thermocouple probe locations are indicated. Figure 2 Indentation marks on tested solder joint Note: Bottom side is the hot side.

5 Thermomigration in lead-free solder joints 15 4 Discussion of observations 4.1 FEM analysis The temperatures at hot and cold side are relatively constant at approximately 160 C and 50 C, respectively, with an error margin of ±1 C. Due to the small size of the solder joints and the associated difficulty in measuring the temperature at the top and bottom of the solder joints, a FEM heat transfer analysis was used. The thermocouple probe locations are used as thermal boundary conditions. There is no heat dissipation to the ambient as the sample is insulated. Temperature dependent material properties, including the aluminum block where the thermocouples were placed, used for the FEM analysis are tabulated in Table 1. The results in Figure 3 show that the top and bottom temperatures of the solder joint are 155 C and 55 C, thus creating a temperature gradient of 1000 C/cm. Table 1 Material properties used in FEM heat transfer analysis Material Density Thermal conductivity Specific heat (kg/m 3 ) (W/m K) (J/kg K) Aluminum Copper SAC405 (NIST, 2004) Note: Properties at 25 C are shown. Figure 3 FEM heat transfer analysis results show highest (bottom) temperature is 155 C while the lowest (top) temperature is 55 C, creating a temperature gradient of 1000 C/cm (see online version for colours)

6 16 M. Abdulhamid, S. Li and C. Basaran 4.2 SEM-EDX observations The experiments are stopped at two hours, 286 hours, 712 hours and 1156 hours. Samples are cross sectioned and analysed using SEM and EDX for microstructural and elemental analyses especially at the hot and cold interfaces. The results were compared to those of isothermally annealed samples of the same stressing time. Isothermal annealing heating took place at 55 C and 170 C, which are almost the same temperatures as the cold and hot side of the test vehicle. Isothermally annealed samples are manufactured the same way as other samples. The Cu 6 Sn 5 IMC at both hot and cold sides of an untested sample is shown in Figure 4. The IMC, which provides bonding between solder joint and Cu plates, is planar and about the same thickness on average at both sides. Using image processing, the average thickness of the IMC layer on top and bottom sides are 6.7 µm and 4.5 µm, respectively. The copper plate/imc and IMC/solder interfaces for both sides are well defined. IMC structure shown in Figure 4 is very well known and studied for solder joints (Tu, 1973). Figure 4 Copper plate/solder joint interface of sample before testing Top Bottom

7 Thermomigration in lead-free solder joints 17 Copper plate/solder joint interface at the colder (top) side after 286 hours, 712 hours and 1156 hours is shown in Figure 5, while the hot (bottom) side is shown in Figure 6. After 286 hours of exposure to the thermal gradient, the hot side (Figure 6) shows a different structure of Cu 6 Sn 5 IMC compared to the cold side (Figure 5). Figure 5 Cold side copper plate-solder joint interface in thermomigration samples showing no development of Cu 3 Sn 286 hours 712 hours 1156 hours Note: Bottom row shows the outline of Cu 6 Sn 5 layer. At the cold side, the copper plate/imc interface is relatively flat, while the IMC/solder joint interface is finger-like as outlined in Figure 5. At both interfaces, a well formed border can be seen separating between plate and IMC, and between IMC and solder. At the hot side, the copper plate/imc joint and IMC/solder joint interface are irregular and wavy as outlined in Figure 6. The Cu 6 Sn 5 IMC thickness is not uniform across the cross section. At some locations, there is no observable IMC layer. Using an image processing software, the average hot side IMC thickness and its ratio to the initial thickness are determined as shown in Table 2. Disintegration of IMC on the hot side is clearly obvious.

8 18 M. Abdulhamid, S. Li and C. Basaran Table 2 Thickness of IMC on the hot side of thermal gradient samples Thermal gradient annealing time (hours) Average hot side IMC thickness (µm) % thickness of initial Figure 6 Hot (bottom) side copper plate-solder joint interface in thermomigration samples showing no development of Cu 3 Sn 286 hours 712 hours 1156 hours Note: Bottom row shows the outline of Cu 6 Sn 5 layer.

9 Thermomigration in lead-free solder joints 19 In Figure 7, copper plate/solder joint interface at the top side after 286 hours, 712 hours and 1156 hours for isothermal annealing at 55 C sample is shown, while at the bottom side for isothermal annealing at 170 C is shown in Figure 8. Evolution of an additional IMC layer is observed between the Cu 6 Sn 5 and Cu plate in 170 C isothermally annealed samples (Figure 8), for the same testing duration. The layer is identified as Cu 3 Sn which is widely reported in solder joint metallurgy literature (Mei et al., 1992). This secondary IMC layer did not form in samples that were subjected to 55 C isothermal annealing and thermal gradient. Figure 7 Top side of isothermal (55 C) samples showing development of only Cu 6 Sn 5 IMC 286h 712h 1156h Figure 8 Bottom side of isothermal (170 C) samples showing development of Cu3Sn IMC between Cu plate and Cu6Sn5 IMC 286h 712h 1156h

10 20 M. Abdulhamid, S. Li and C. Basaran After 1156 hours, the samples that were subjected to thermal gradient Cu 6 Sn 5 IMC structure at the cold (top) side becomes finger-like, while the hot side IMC becomes thinner (Table 2). The 170 C isothermal samples show an increase in thickness of Cu 6 Sn 5 and Cu 3 Sn IMC which is not evident in thermal gradient samples. The Cu 6 Sn 5 layer thickness is about 2.3 times thicker than Cu 3 Sn, as shown in Table 3. Table 3 IMC thickness for isothermal annealed samples Isothermal annealing time (hours) Average IMC thickness Cu 6 Sn 5 (µm) Cu 3 Sn (µm) Ratio After reflow Nano indentation testing In this study, MTS nanoindenter was used for hardness and modulus measurements which details are given in (Oliver and Pharr, 1992). In this experiment, the hardness measurement is based on a prescribed maximum load of 250 mn after considering the surface area size and number of indentation that needed to be done. Before achieving the prescribed load, the measurements were carried out at four different loads. As shown in Figure 9, at maximum load, the hardness has reached an asymptotic value which indicates the actual hardness of the material. A typical hardness measurement for one row of indentation points are shown in Figure 9. The numerical data and statistical analysis for Figure 9 are presented in the appendix. The mean hardness for every sample versus distance from hot side is plotted in Figure 10. Numerical data are available in the appendix. The average surface hardness (Figure 10) from nano indentation tests shows that in samples subjected to thermal gradient (TG), hardness increases from the hot side to the cold side at a relatively constant rate. Except for samples that are under TG for two hours, the hardness values for other samples are within the same range across the surface. The hardness value for samples which experienced TG for two hours shows a higher hardness value than the other near the hot side, but about the same value as measurement nears the cold side. In all TG cases, the hardness value near the cold side is slightly lower than the untested sample. The average hardness for the as-flowed (untested) samples is shown for reference. The average hardness for isothermally annealed (IT) samples is relatively constant, and lower than the TG samples. Structural changes for the isothermal samples all occurred within the first test interval of 286 hours.

11 Thermomigration in lead-free solder joints 21 Figure 9 A plot of hardness vs. indentation depth for specimen 1 for TG experiment after 712 hours at a normalised distance 0.95 from hot side (see online version for colours) Hardness of Specimen 1 for TG 712h Sample at Distance 0.95 Hardness (GPa) Indentation Depth (nm) Figure 10 Average measured surface hardness of thermomigration and isothermal samples across the solder height (TG: thermal gradient, IT: isothermally annealing) (see online version for colours) Average Measured Hardness for Isothermal and Thermal Gradient Samples Hardness (GPa) As flowed TG 286 TG 712 TG 1156 IT 286 IT 712 IT 1156 TG Normalized Distance from Hot Side

12 22 M. Abdulhamid, S. Li and C. Basaran 4.4 Vacancy migration simulations Ye et al. (2003d) developed a general mechanically coupled vacancy migration model based on Kircheim (1992, 1993) formulation, C v = qv + G (1) t where the vacancy flux due to diffusion driving forces is given by DC v v * DC v v * T DC v v qv = Dv Cv Zve ( ρ j) Qv + ( fω) σ kt kt T kt Combining (1) and (2), C v DvC v * DvCv * T DvC v = Dv Cv Zve ( ρ j) Qv + ( fω) σ + G t kt kt T kt (2) (3) where G is the vacancy generation rate [Sarychev and Zhitnikov (1999)], C G = v C τ s ve (4) C ve is the thermodynamics equilibrium vacancy concentration, C ve ( 1 f ) Ωσ = Cv0 exp (5) kt σ is hydrostatic or spherical part of the stress tensor defined as, trace( σ ij ) σ = (6) 3 D v is vacancy diffusivity rate (Ye at al., 2003d), D v Da = 5 (7) 3 10 The rest of the parameters are as follows C v vacancy concentration D a atomic diffusivity rate * Z v vacancy effective charge number e ρ j electron charge metal resistivity current density vector

13 f Ω k T Thermomigration in lead-free solder joints 23 vacancy relaxation vector atomic volume Boltzmann constant absolute temperature C v0 equilibrium vacancy concentration in absence of stress τ s characteristic vacancy generation/annihilation time. The derivation of equation (3) is explained in detail in (Ye et al., 2003d). In equation (3), the first term in the square bracket represents the driving force due to concentration gradient, second term represents the driving force due to electrical current, third term due to TG, and the final term is due to back stress. Since in this experiment, the samples are not subjected to electrical current, the second term can be dropped. By defining relative concentration, C C v /C v0, and dropping the electromigration driving force term, equation (3) can be reduced to C DC v * T DC v G = Dv C Qv + ( fω) σ + (8) t kt T kt Cv0 where initially, C v = C v0. Since direction of vacancy flux is in the opposite direction of atomic flux, q v = q a, it follows that Q * v = Q * a in absence of electromigration (Kirchheim, 1993). Vacancy is considered as a substitutional species which has a smaller relaxed volume than that of an atom at a lattice site. When an exchange of vacancy and atom occurs at a lattice site, a local volumetric strain occurs. This local volumetric strain is also produced as vacancy is generated or annihilated at vacancy source or sink, respectively. The vacancy source and sink can be at grain boundary or at a dislocation (Kirchheim, 1993). Sarychev and Zhitnikov (1999) proposed that the vacancy diffusion causes volumetric strain which has two parts. One part of the volumetric strain is due to vacancy flux divergence, and the other is due to vacancy generation. These strains are time dependent and are expressed in term of strain rate. For vacancy flux divergence, 1 ε = f Ω q δ 3 d ij v ij (9) while for vacancy generation, g 1 εij = ( 1 f ) ΩGδ ij (10) 3 The total volumetric strain due to vacancy migration is, v d g εij = εij + εij v ε =Ω f q+ f ( 1 ) G (11)

14 24 M. Abdulhamid, S. Li and C. Basaran Total strain in the material will be sum of the three components, ε = ε + ε + ε (12) total mech therm v ij ij ij ij where the first term is strain due to mechanical loading, second is due to thermal loading, and the last one is due vacancy migration. Quasi-static mechanical equilibrium equation is given by, σ ij x j = 0 where σ ij is the stress tensor and x j is jth coordinates. The stress is related to strain according to σ = ε (14) K stiffness where K stiffness is the tangential stiffness matrix. In this simulation, the material model developed by Lin and Basaran (2005) was used. By solving equations (8), (11) and (13), the vacancy concentration C v can be determined. Parameters used to solve these equations are listed in Table 4. Table 4 Parameters used in simulation Parameter Value Reference D a cal 8.2exp RT cm/s2 Ye et al. (2003d), Singh and Ohring (1984) C v cm 3 Ye et al. (2005) f 0.2 Bassman (2000) Ω cm 3 Ye et al. (2005) * Q a 0.23 ev Chuang and Liu (2006) Τ s s Sarychev et al. (2000) E 57.7 GPa Hong (1998) α 18.9 ppm Chen et al. (2003) A plot of relative vacancy concentration across the solder joint height for different stressing times is shown in Figure 11. The plot shows that near the hot side, the relative vacancy concentration reaches a constant value about 1.09, and about one near the cold side, after 100 hours. This is evident since the relative vacancy at 300 hours is the same as that of at 100 hours. Near the hot side, as time progresses, the vacancy concentration increases to a maximum value about 9% above its initial concentration. Near the cold side, the vacancy concentration initially increases about 2% then decreases to its initial value as time progresses. A plot of relative vacancy concentration as function of time, in Figure 12, shows the vacancy concentration reaches a steady state after 20 hours. Near the cold side, the vacancy concentration peaks at about 2% above its initial value after two hours before declining to its initial value after 20 hours. (13)

15 5 Discussion Thermomigration in lead-free solder joints 25 The two major microstructural differences observed between thermomigration and isothermal annealing experiments are the absence of Cu 3 Sn IMC at both cold and hot sides in thermomigration samples, and the thinning of Cu 6 Sn 5 IMC layer on the hot (155 C) side in thermomigration samples. The thinning of as-flowed Cu 6 Sn 5 IMC layer at the hot side is due its disintegration to 6 Cu and 5 Sn atoms under very large TG according to equation (15). Cu6Sn5 6Cu + 5Sn Figure 11 Relative vacancy concentration across the solder cross section as a function of time (see online version for colours) (15) Relative Vacancy Concentration as a Function of Distance Relative vacancy concentration hours 10 hours 100 hours 300 hours Initial Normalized distance from hot side

16 26 M. Abdulhamid, S. Li and C. Basaran Figure 12 Relative vacancy concentration near hot and cold side as a function of stressing time (see online version for colours) Relative Vacancy Concentration as a Function of Stressing Time Relative concentration distance (near cold side) distance (near hot side) Stressing time (hours) The Soret effect from TG segregates the Cu 6 Sn 5 IMC to Sn- and Cu-rich layers, as shown in Figure 13(a). The Cu atoms from Cu-rich layer drift to the cold side under the TG force. This is due to the fact that diffusivity of Cu is greater than that of Sn. A similar result was observed by Ding et al. (2005, 2006) under electromigration at 150 C isothermal temperature. They conclude that Cu atoms from Cu-UBM and Cu 6 Sn 5 dissolutions drift to bulk solder under electromigration driving force. However, in their experiment, they did not isolate thermomigration from electromigration. The dissolution of Cu 6 Sn 5 is due to TG and not electromigration because both Sn and Cu atoms are exposed to the same temperature but only Cu atom diffuses. At any temperature, Cu atom diffuses orders of magnitude faster than Sn. Since a Cu atom is lighter (atomic mass of 63.5 g/mol and atomic radius of 1350 Å) than the Sn atom (118.7 g/mol and 1450 Å), and being the dominant diffusion species (Tu, 1973), Cu atoms move faster than Sn atoms to the cold side under the TG driving force. The diffusivity rates, D Cu-in-Sn and D Sn-in-Sn at 20 C are m 2 /s and m 2 /s, respectively, at 190 C, D Cu-in-Sn and D Sn-in-Sn are m 2 /s and m 2 /s (Mei, 1992). Cu diffusivity is four order of magnitude faster than Sn diffusivity at 20 C, and three at 190 C. This difference in migration speed produces segregation effect in which Cu is seen to migrate to the cold side while Sn stay relatively stationary in the original site, which is consistent with previous studies (Huang et al., 2006; Schroerschwarz, 1971). The apparent movement of Cu to the cold side and accumulation of Sn near hot side is caused by the different rate of diffusivities. Under TG force, both atoms move the cold

17 Thermomigration in lead-free solder joints 27 side. Since Cu has a much higher rate of diffusivity than Sn, Cu will move faster than Sn to the cold side. In the long run, this difference in diffusivity rate segregates Cu and Sn atoms. High concentration of Cu can be seen on the cold side, while Sn on the hot side. While Cu atoms are seen to migrate to the cold side, Sn atoms from the disintegration react with Cu atoms from the Cu plate to form a new thin layer of Cu 6 Sn 5 as shown in Figure 13(c). This new Cu 6 Sn 5 IMC layer maintains the bond between the Cu plate and the solder joint. Since the Cu plate provides unlimited Cu atoms, there will always be a thin layer of Cu 6 Sn 5 IMC layer. The layer thickness is governed by the rate of dissolution of old and formation of new Cu 6 Sn 5 IMC. In this experiment, the thickness is about 3.5 µm. If supply of Cu atom is limited, such in the case of a thin Cu-UBM, there will not be a Cu 6 Sn 5 after some period of time. This depletion of Cu atom from Cu-UBM and Cu dissolution from Cu 6 Sn 5 creates a gap between UBM and solder bulk thus creating reliability issues. If IMC layer is too thin, there will be no adhesion between Cu plate and solder bulk (Tu and Zeng, 2001), while too thick a layer will decrease fracture toughness (Pratt et al., 1996). Since Sn is highly reactive with Cu, a normal practice in the industry is to have a Ni/Au diffusion barrier coating on the Cu pad before Sn rich solder joint is attached. Ni/Au coating not only helps in preventing oxidation and corrosion, but also serves as a diffusion barrier (Fu et al., 2006). Figure 13 The process of old Cu 6 Sn 5 layer disintegration into new layer of Cu 6 Sn 5 (a) (b) (c)

18 28 M. Abdulhamid, S. Li and C. Basaran In the isothermal annealing case, Cu 3 Sn is formed between Cu plate and Cu 6 Sn 5 layer, which is similar to earlier studies (Tu, 1982; Yoon, 2004) for isothermal aging between 100 C to 350 C. Deng et al. (2003) suggest the Cu 3 Sn growth is due to the reaction of Cu with Cu 6 Sn 5, according to equation (16), while Zeng et al. (2005) argued that the growth is due to either the reaction of Sn and Cu, or decomposition of Cu 6 Sn 5, according to equation (17). The reaction requires one Sn and three Cu atoms, while the decomposition of Cu 6 Sn 5 releases two Cu 3 Sn molecules and three Sn atoms (Zeng et al., 2005). 9Cu + Cu6Sn 5 5Cu3Sn (16) Cu6Sn 5 2Cu3Sn + 3Sn (17) At the hot side in the TG case, there is no observable Cu 3 Sn IMC. Even though the supply of both Cu from the plate and Sn from the solder is abundant, the required mass concentration combination is not achieved to form Cu 3 Sn in the TG samples. At the hot end, where temperature is below 180 C, the required Cu mass concentration of between 39.1% and 55.2% (Saunders and Miodownik, 1990) to form Cu 6 Sn 5 is achieved. The required Cu mass concentration to form Cu 3 Sn is between 60.5% and 72.3% (Saunders and Miodownik, 1990). This requirement is achieved in isothermal samples but is not achieved in TG samples indicating that, while some Cu atoms reacted with Sn atoms to form a new Cu 6 Sn 5 layer as discussed in an earlier paragraph, the other drifted to the cold side under TG force. While the cold (55 C) side of the thermomigration sample is not favourable for Cu 3 Sn IMC formation, the excess Cu coming from the hot side reacts with existing Cu 6 Sn 5 IMC to form finger-like Cu 6 Sn 5. An EDX analysis near the cold side (points four to six) after 712 hours shows high concentration of Cu as shown in Table 5 and Figure 14. Under TG, the hardness shows an increasing trend from hot to cold side, while in the isothermal case, hardness remains relatively constant. The hardness degradation from cold side to hot side in the thermomigration samples is attributed to combination of vacancy migration and Sn-grain coarsening from cold to hot side. Table 5 Element analysis near the cold interface after 712 hours Point Sn (%wt) Ag (%wt) Cu (%wt)

19 Thermomigration in lead-free solder joints 29 Figure 14 SEM image of the cold interface after 712 hours Note: Points 1 to 3 are identified as Cu 6 Sn 5 IMC. Point 4 to 6 show increase in Cu concentration as points are closer to the interface. Grain size increase and hardness degradation from cold to hot side can be explained using Hall-Petch relation for hardness (Tjong and Chen, 2004), H H k d 1/2 v = O + H (18) where H O and k H are material constants, and d is grain diameter. Smaller grain size at the cold side means more grain boundaries, which act as dislocation motion barriers. As dislocation slip motion is impeded, stresses required to continue the slip process increase (Hall, 1951), as a result, the material has a greater hardness. For isothermal samples, the grain size is uniform across the solder joint according to this relation, and accordingly there is no gradient of hardness. Lead-free solder alloy, including SAC, has been shown to follow Hall-Petch relation (Siviour et al., 2005). For SnPb solder, nanoindetation results from Ye et al. (2004) show that the mechanical properties are in agreement with the Hall-Petch relation. They observed Pb-grain coarsening after thermomechanical loading, and found that solder alloy properties degrade from substrate (cold) to chip (hot) side. Investigation of Pb-grain size shows that the grain is larger at the hot side than at the cold side as shown in Figure 15. Vacancy migration is a result of Cu and Sn atoms movement to the cold side. When these atoms migrate, they fill the vacant lattice sites or vacancies in the cold side by substitution mechanism. Comparing Figure 10 and Figure 11, it can be observed that near the hot side where vacancy concentration is high, the hardness is low. In other words, near the hot side, the solder is more porous than the cold side. From Figure 12, it can be observed that after 20 hours of stressing, the vacancy concentration at both hot and cold side, reaches an asymptotic value. This explains why the hardness for samples stressed above 20 hours in Figure 10 is within the same range.

20 30 M. Abdulhamid, S. Li and C. Basaran Figure 15 Pb-grain coarsening reported by Ye at al. (2004) (see online version for colours) Note: At the chip (hot) side, Pb grain size is larger compared to the substrate (cold) side. 6 Conclusions The microstructural and mechanical properties of lead-free solder joint/copper pad interface were studied under a temperature gradient of 1000 C/cm. The two major microstructural differences between thermomigration and IT samples are the lack of Cu 3 Sn IMC layer at both hot and cold side, and the thinning of Cu 6 Sn 5 IMC layer at the hot side under to TG driving force. In the thermomigration samples, the thinning of Cu 6 Sn 5 layer is a result of its disintegration, while the absence of Cu 3 Sn IMC layer is a result of insufficient Cu mass concentration to form Cu 3 Sn layer. The samples form only Cu 6 Sn 5 IMC and show more Cu concentration near the cold side which resulted in a well defined Cu 6 Sn 5 /solder interface. In the isothermal annealing case, Cu 3 Sn layer is formed between the Cu plate and Cu 6 Sn 5 layer. The Cu 6 Sn 5 layer thickness is about 2.3 that of Cu 3 Sn layer over time, suggesting that the growth rate of Cu 6 Sn 5 is faster than that of Cu 3 Sn. Under TG driving force, the hardness degrades from cold to hot side which could be attributed to vacancy migration and Sn-grain coarsening in the same direction. Vacancy concentration increases from cold to hot side and reaches steady state after 20 hours. Hardness degradation was not observed across solder joint for isothermal aging samples, which indicates uniformity in Sn-grain size. Acknowledgements This project has been partly sponsored by US Navy Office of Naval Research Advanced Electrical Power Systems under the direction of Terry Ericsen.

21 Thermomigration in lead-free solder joints 31 References Basaran, C., Lin, M. and Ye, H. (2003) A thermodynamic model for electrical current induced damage, International Journal of Solids and Structures, Vol. 40, pp Basaran, C., Ye, H., Hopkins, D.C., Frear, D. and Lin, J.K. (2005) Failure modes of flip chip solder joints under high electric current density, Journal of Electronic Packaging, Vol. 127, pp Bassman, L.C. (2000) Modelling of stress-mediated self-diffusion in polycrystalline solids, PhD, Stanford University. Chen, Z., Shi, Y., Xia, Z. and Yan, Y. (2003) Properties of lead-free solder SnAgCu containing minute amounts of rare earth, Journal of Electronic Materials, Vol. 32, pp Chuang, Y.C. and Liu, C.Y. (2006) Thermomigration in eutectic SnPb alloy, Applied Physics Letters, Vol. 88, pp Deng, X., Piotrowski, G., Williams, J.J. and Chawla, N. (2003) Influence of initial morphology and thickness of Cu6Sn5 and Cu3Sn intermetallics on growth and evolution during thermal aging of Sn-Ag solder/cu joints, Journal of Electronic Materials, Vol. 32, pp Ding, M., Wang, G., Chao, B., Ho, P.S., Su, P. and Uehling, T. (2006) Effect of contact metallisation on electromigration reliability of Pb-free solder joints, Journal of Applied Physics, Vol. 99, pp Ding, M., Wang, G., Chao, B., Ho, P.S., Su, P., Uehling, T. and Wontor, D. (2005) A study of electromigration failure in Pb-free solder joints, in 43rd IEEE International Reliability Physics Symposium Proceedings, pp Fu, R., Liu, L., Liu, D. and Zhang, T-Y. (2006) In situ formation of Cu-Sn-Ni intermetallic nanolayer as a diffusion barrier in preplated lead frames, Applied Physics Letters, Vol. 89, pp Hall, E.O. (1951) The deformation and aging of mild steel: III. Discussion of results, in Proceedings of the Physical Society, London, Vol. B64, pp Hong, B.Z. (1998) Thermal fatigue analysis of a CBGA package with lead-free solder fillets, in Thermal and Thermomechanical Phenomena in Electronic Systems, ITHERM 98, The Sixth Intersociety Conference, pp Huang, A.T., Gusak, A.M., Tu, K.N. and Lai, Y-S. (2006) Thermomigration in SnPb composite flip chip solder joints, Applied Physics Letters, Vol. 88, pp JEDEC Solid State Association and IPC (2004) Moisture/Reflow Sensitivity Classification for Nonhermetic Solid State Surface Mount Devices. Kirchheim, R. (1992) Stress and electromigration in Al-lines of integrated circuits, Acta Metallurgica et Materialia, Vol. 40, pp Kirchheim, R. (1993) Modeling electromigration and induced stresses in aluminum lines, in Materials Research Society Symposium Proceedings, San Francisco, pp Lin, M. and Basaran, C. Electromigration induced stress analysis using fully coupled mechanical-diffusion equations with nonlinear material properties, Computational Materials Science, Vol. 34, pp Mei, Z., Sunwoo, A.J. and Morris, Jr., J.W. (1992) Analysis of low-temperature intermetallic growth in copper-tin diffusion couples, Metallurgical Transactions A: Physical Metallurgy and Materials Science, Vol. 23A, pp NIST (2004) Review and Analysis of Lead-Free Solder Material Properties, Metallurgy Division of Material Science and Engineering Laboratory, NIST. Available online at: Oliver, W.C. and Pharr, G.M. An improved technique for determining hardness and elastic modulus using load and displacement sensing indentation experiments, Journal of Materials Research, Vol. 7, p.1564.

22 32 M. Abdulhamid, S. Li and C. Basaran Platten, J.K. (2006) The Soret effect: a review of recent experimental results, Journal of Applied Mechanics Transactions of the ASME, Vol. 73, pp Pratt, R.E., Stromswold, E.I. and Quesnel, D.J. (1996) Effect of solid-state intermetallic growth on the fracture toughness of Cu/63Sn-37Pb solder joints, IEEE Transactions on Components, Packaging, and Manufacturing Technology, Part A, Vol. 19, pp Roush, W. and Jaspal, J. (1982) Thermomigration in lead-indium solder, in Electronic Components Conference, San Diego, CA, pp Sarychev, M.E. and Zhitnikov, Y.V. (1999) General model for mechanical stress evolution during electromigration, Journal of Applied Physics, Vol. 86, p Sarychev, M.E., Zhitnikov, Y.V., Borucki, L., Liu, C.L. and Makhviladze, T.M. (2000) A new general model for mechanical stress evolution during electromigration, Thin Solid Films, Vol. 365, pp Saunders, N. and Miodownik, A.P. (1990) The Cu-Sn (copper-tin) system, Bulletin of Alloy Phase Diagrams, Vol. 11, pp Schroerschwarz, R. and Heitkamp, D. (1971) Thermotransport of substitutional impurities in copper, Physica Status Solidi (B), Vol. 45, pp Singh, P. and Ohring, M. (1984) Tracer study of diffusion and electromigration in thin tin films, Journal of Applied Physics, Vol. 56, pp Siviour, C.R., Walley, S.M., Proud, W.G. and Field, J.E. (2005) Mechanical properties of SnPb and lead-free solders at high rates of strain, Journal of Physics D: Applied Physics, Vol. 38, pp Soret, C. (1879) Sur l état d équilibre que prend au point de vue de sa concentration une dissolution saline primitivement homohéne dont deux parties sont portées à des températures différentes, Archives des Sciences Physiques et Naturelles, Vol. 2, pp Tjong, S.C. and Chen, H. (2004) Nanocrystalline materials and coatings, Materials Science and Engineering: R: Reports, Vol. 45, pp Tu, K.N. (1973) Interdiffusion and reaction in bimetallic Cu-Sn thin films, Acta Metallurgica, Vol. 21, pp Tu, K.N. and Thompson, R.D. (1982) Kinetics of interfacial reaction in bimetallic Cu-Sn thin films, Acta Metallurgica, Vol. 30, pp Tu, K.N. and Zeng, K. (2001) Tin-lead (SnPb) solder reaction in flip chip technology, Materials Science and Engineering: R: Reports, Vol. 34, pp Ye, H. (2004) Mechanical behaviour of microelectronics and power electronics solder joints under high current density: Analytical modeling and experimental investigation, PhD Dissertation, State University of New York, Buffalo. Ye, H., Basaran, C. and Hopkins, D.C. (2003a) Damage mechanics of microelectronics solder joints under high current densities, International Journal of Solids and Structures, Vol. 40, pp Ye, H., Basaran, C. and Hopkins, D.C. (2003b) Measurement of high electrical current density effects in solder joints, Microelectronics Reliability, Vol. 43, pp Ye, H., Basaran, C. and Hopkins, D.C. (2003c) Mechanical degradation of microelectronics solder joints under current stressing, International Journal of Solids and Structures, Vol. 40, pp Ye, H., Basaran, C. and Hopkins, D.C. (2003d) Numerical simulation of stress evolution during electromigration in IC interconnect lines, IEEE Transactions on Components and Packaging Technologies, Vol. 26, pp Ye, H., Basaran, C. and Hopkins, D.C. (2003e) Thermomigration in Pb-Sn solder joints under joule heating during electric current stressing, Applied Physics Letters, Vol. 82, pp

23 Thermomigration in lead-free solder joints 33 Ye, H., Basaran, C. and Hopkins, D.C. (2004) Pb phase coarsening in eutectic Pb/Sn flip chip solder joints under electric current stressing, International Journal of Solids and Structures, Vol. 41, pp Ye, H., Basaran, C., Hopkins, D.C. and Lin, M. (2005) Modeling deformation in microelectronics BGA solder joints under high current density Part I Simulation and testing, in Proceedings of the 55 th Electronic Components and Technology Conference, pp Yoon, J-W., Lee, Y-H., Kim, D-G., Kang, H-B. Suh, S-J., Yang, C-W., Lee, C-B., Jung, J-M., Yoo, C-S. and Jung, S-B. (2004) Intermetallic compound layer growth at the interface between Sn-Cu-Ni solder and Cu substrate, Journal of Alloys and Compounds, Vol. 381, pp Zeng, K., Stierman, R., Chiu, T-C., Edwards, D., Ano, K. and Tu, K.N. (2005) Kirkendall void formation in eutectic SnPb solder joints on bare Cu and its effect on joint reliability, Journal of Applied Physics, Vol. 97, pp Appendix Table A1 Numerical data and statistical analysis for Figure 9 Indentation number Data at prescribed maximum load Hardness (GPa) Modulus (GPa) Depth (nm) Load (mn) Mean Std. dev % COV

24 34 M. Abdulhamid, S. Li and C. Basaran Appendix (continued) Table A2 Average measured hardness from nano indentation tests