Joint Microstructure

The macroscopic behavior of a joint depends on the initial microstructure, and its evolution during loading. A fine microstructure is desirable due to its superior creep resistance. Cracks have also been observed to initiate faster in coarser regions (Wild, 1975; Mei and Morris, 1992). Shen et al (2001) argue that the coarsened region is not weaker per se, but is subjected to earlier damage due to higher plastic strains.

Initial Microstructure

The initial state of a joint is determined by the soldering process and the subsequent cooling rate. The latter depends on the size, composition and configuration of the board, as well as the soldering process.

Slow cooling rates (Conrad et al., 1999) for Pb/Sn eutectic solder result in stable, alternating lamellae of the tin and lead-rich phases, parallel to the direction of growth, with the fineness of the microstructure increasing with faster cooling (Lau, 1991). Faster cooling rates and the presence of impurities (Jackson, 2002) lead to increasingly finer globular degenerate structures, with quenching leading to a dispersion of the lead-rich phase in a tin-rich matrix (Mei and Morris, 1992).

The excellent solvent properties of molten solder (Ross, 1991) imply that surroundings of the joint determine the presence of impurities, as well as the extent of oxidation. Instabilities such as unfavorably oriented growth and local variation in cooling rates may cause a breakdown in the long-range structure and short-range colonies, with homogenous phase distributions within, may arise. Colonies are separated from each other by a distinct colony boundary which is delimited by a narrow region of enlarged, equiaxed particles (Tribula et al., 1989; Tribula and Morris, 1990).

The microstructure of 60/40 tin-lead solder is similar to that of the eutectic alloy, but coarser and with less distinct boundaries (Summers and Morris, 1990), with pro-eutectic lead rich particles containing plate-shaped tin-rich precipitates dispersed in the eutectic structure (Tribula et al. 1989).

Intermetallics

During soldering, molten solder on the substrate, typically copper, instantly forms intermetallic copper-tin compounds and the layer rapidly thickens to a few microns. This forms the bonding layer between the solder and the copper (So et al., 1997).

The intermetallic consists of two phases, a thin layer of Cu₃Sn (e-phase) at the copper pad, and rod-like hexagonal projections of Cu₆Sn₅ (h-phase) into the solder in contact with the tin. The rods can break off into the molten solder, appearing as intermetallic particles or whiskers in bulk solder (Tribula et al., 1989). Formation of intermetallics may be suppressed in the presence of additional elements. Cooling rates along with the metallization of the connecting pads determines the nature, quantity and distribution of the intermetallics formed.

For Pb/Sn solder, intermetallic growth has also been observed to continue with time depending on the temperature, obeying

where d is the layer thickness, D is the interdiffusion constant and t is the aging time. The dependence of layer growth in terms of temperature can be given by the Arrhenius equation

where D₀ is the interdiffusion constant, Q is the activation energy for the growth of the intermetallic layer, k is the Boltzmann constant and T is the absolute temperature.

Wassink (1989) reported growth even at room temperature, while So et al. (1997) cite results from literature indicating that the intermetallic thickness is unaffected by up to 13 weeks of heating at 70^oC, while heating at 170^oC leads to thickening of both phases, as well as conversion of the h-phase to the e-phase. Intermetallic growth under thermal cycling with no applied mechanical loads has been reported by Pang et al. (2001).

Microstructure Evolution

The as-cast degenerate microstructure of Pb/Sn joints (Mei et al., 1991; Mei and Morris,1992; Conrad et al., 1999) obtained in the presence of impurities (Jackson, 2002), is unstable (Tribula et al., 1989; Cutiongco et al., 1990; Mei et al., 1991) with respect to coarsening of the lead-rich region due to diffusion of vacancies (Vianco et al., 1999) under aging (Cutiongco et al., 1990; Ross, 1991), thermal cycling (Pang et al., 2001) and especially thermomechanical loading (Hacke et al., 1997, 1998) where the number of vacancies is higher. Equilibrium is attained under isothermal annealing at 100^oC after a few days, or a few months at room temperature (Cutiongco et al., 1990; Ross, 1991). Solomon (1986) did not find significant microstructural changes after fatigue at 35^oC for a relatively coarse as-cast structure while recrystallization and grain growth were observed by Wild (1975), especially for long-duration high-temperature tests. Summers and Morris (1990) noted that strains were more homogenous and the coarsening was less pronounced in 60/40 tin-lead solder as compared to the eutectic solder.

Ag/Sn joints, with less than 5% intermetallic Ag₃Sn by volume, show coarsening under aging at higher temperatures only (Yang et al., 1994; 1995). However, a large number of intermetallic particles, while increasing the creep resistance also result in increased sites for microcrack initiation (Igoshev and Kleiman, 2000).

Under shear deformation, inhomogeneous deformations observed in eutectic and near‑eutectic solders have been observed to lead to a band of coarsened material slightly offset from the interface. Tribula et al. (1989) ascribed the growth of the band due to the relatively soft coarsened material acting like a Type II crack creating a narrow band of high plastic deformations, with recrystallization leading to propagation at the tip. However, Shen et al. (2001) argue that locally coarsened material may not be weaker than the surrounding region, but is subject to early damage initiation due to accumulation of plasticity.

Damage Accumulation

Cracks were noted in all cyclic loading studies within the coarsened band, and were observed to merge and form thermal fatigue cracks. Faster initiation and propagation of cracks in coarser microstructures has been observed by Wild (1975) and Mei and Morris (1992). Attarwala et al. (1992) provided experimental confirmation for creep-fatigue interactions under thermomechanical cycling, with the observation of fatigue striations related to crack growth, from cracks initiated mainly in the harder tin region, as well as voids. Mechanistic models for crack initiation and propagation have been proposed by Wong and Helling (1990) and Stone (1990) for crack initiation due to grain-boundary sliding, thermal nucleation at triple points, and subsequent propagation.

Zubelewicz et al. (1989) reported transgranular and mixed fracture modes for high strain ranges, and intergranular for low strain ranges. Tribula et al. (1989) reported formation of coarsened bands under shear loading for unidirectional loading and thermomechanical cycling leading to strain concentrations and thus, crack growth. The band was seen to follow colony boundaries at low strain rates, while at high rates the band was parallel to the direction of shear. Similar results have been obtained for thermomechanical cycling by Frear et al. (1995). Under tensile cycling, failure was obtained at the brittle intermetallic. Failure at this site may also be expected under impact loads. Lee and Stone (1992) reported shear localization and transgranular failure with a dimpled fracture surface for high strain rates, and failure by cavitation at low strain rates due to grain boundary sliding for tensile loading. Under tension-compression fatigue, low frequency and high strain range lead to numerous intergranular cracks at the surface, which propagated along grain boundaries to the interior, while high frequencies at low strains led to transgranular failure.

Cutiongco et al. (1990) reported increased fatigue life after heating at 150^oC for up to a week, with a subsequent leveling-off. Vaynman et al. (1998) reported increased ductility and fatigue life for joints aged for seven years. Mei and Morris (1992) showed that quenched joints had about twice the fatigue lifetime for isothermal cycling as compared to furnace cooled joints. Chan et al. (1997) showed a decrease in fatigue life with intermetallic thickness. Summers and Morris (1990) reported that a slight change in microstructure, such as that between 63/37 and 60/40 lead-tin solder, could lead to substantial differences in fatigue lives.

Microstructure and its effects on solder behavior, including constitutive response as well as damage accumulation must be considered in design of accelerated tests, to ensure that the fundamental mechanisms of deformation and damage accumulation being modeled are the one that is expected under field conditions. Frear et al. (1995) have provided guidelines for design of thermomechanical tests for Pb/Sn solder.