The methods of both experiments and phase field simulation were combined to investigate the formation process of Sn-rich phase microstructure on SAC305 balls and powder products. Results show that experiments consist well with the simulation as discussed below.
3.1. Microstructural Characteristics by Experimental Methods
Morphologies or patterns are shaped by the complex dynamical evolution of the solid-liquid interface and are both intricate and varied.
Figure 2 shows the microstructure of a BGA solder ball. It can be seen from
Figure 2a,b that many continuous humps distributed on a solder ball surface and there were both smooth areas (zone A) and rough areas (zone B). The micrographs clearly illustrate the diffusion induced phase separation in the considered Sn-Ag-Cu system. Details of the BGA solder ball are shown at higher magnification in
Figure 2c,d imaged by SEM. As a rule, worm-like structure which is typically observed in spinodal decomposition was found on the solder ball surface. Generally, non-bright humps area had two kinds of shape form: firstly, rod humps regularly precipitated with long sides connected with each other as shown in the dashed line box in
Figure 2c; secondly, nearly round shape distributed randomly as shown in the dotted circle in
Figure 2b. These matt-like humps distributed discretely were nearly oval in cross-section with a diameter of about 1~10 μm. Moreover, most of worm-like humps on a solder ball surface were rod-like and with length of about 20 μm and short edge 5 μm; the periodicity as a mean distance between two neighboring structures ranged from 2 to 5 μm. Additionally, many inter-metallic compound (hereinafter referred to as IMC) Ag
3Sn particles dispersed on the solder ball surface. They precipitated along humps like “necklace” as the dashed curve in
Figure 2d indicates. The regular distribution of Ag
3Sn indicates that the approximate spinodal decomposed patterns form firstly and then the Ag
3Sn precipitates afterwards. It was reported that cracks on the surfaces of solder powder balls would initiate more easily than that in solder balls since there were more trenches on the surface of the latter [
6]. Otherwise, the matt-like morphology on the ball surface can cause blurred vision during BGA ball mounting. Therefore, it is quite necessary to study the formation mechanism of such morphology so as to control the surface structure of BGA balls.
EDS experiments were carried out in order to study the composition of humps, as
Figure 3 shows. The EDS scanning experiment results show that the composition of humps was pure Sn (100 wt. %), as shown in
Figure 3b. The matrix of the solder ball contained three elements Sn, Ag and Cu, and the content (96.6 Sn 2.88 Ag 0.468 Cu) was roughly the same as products of solder balls, as shown in
Figure 3c. As can be seen, two phases of different composition emerged, where the hump areas were the Sn-rich phase and the other was the matrix area of solder ball. The Sn content on the solder ball surface was 93.44 wt. %. The EDS results indicated that it was the component Sn diffused out of the solder ball surface to form the Sn-rich phase up to the “matt-like” morphology. In conclusion, the Sn-rich phase distributed unevenly in the SAC305 alloy matrix of the solder interconnect during thermal aging, which formed the morphology of coexistence of smooth and rough areas as
Figure 2b presents.
The elemental SEM-EDS mapping was performed in
Figure 4 to further study the composition distribution of SAC305 solder ball. As seen, Sn contrast (in green) was higher in the Sn-rich phase than that in the alloy matrix, while its intensity decreased in the Ag
3Sn IMC regions. The distribution of different shades of red, yellow and green indicated that the originally mix of Sn, Ag and Cu was superseded by islands of high Sn concentration. These form as a result of a diffusion driven separation of both phases known as coarsening [
15]. Ag (in yellow) widely distributed on the solder matrix with obvious edges compared with Sn and Cu, which can be deduced to be the reason that many Ag
3Sn formed on the solder ball surface. These results also indicated that Ag was consumed in the formation of Ag
3Sn IMC. The presence of Cu (in red) was not obvious due to its relatively low content. Therefore, it is confirmed that the formation of the Sn-rich phase was the diffusion of Sn from low concentration to high concentration. In other words, Sn separated from the matrix to form Sn-rich phase with a certain composition higher than that of the solder ball. The alloy spontaneously disintegrates into Sn-rich poor phase and matrix [
16]. Moreover, there was no obvious phase interface between Sn-rich phase and the SAC305 matrix. To sum up, this micromorphological change was a result of a diffusion process of the component Sn, also known as spinodal decomposition, followed by coarsening through phase separation. In other words, spinodal decomposition was realized through spontaneous growth of fluctuations of Sn composition. Consequently, this process of Sn phase separation led to worm-like stochastically orientated patterns. Heat treatments during production yielded a fluctuation of component distribution, formed during heating, with evidence supporting its formation to be via spinodal decomposition [
17]. Such types of phase separation can be accelerated considerably if the homologous temperatures are high [
18]. For example, it was reported by Kaban et al. [
19] that the surface tension of the investigated Sn-Ag-Cu alloys strongly depends on Sn concentration. The surface tension decreases rapidly with small additions of Sn and varies weakly with composition in the Sn-rich region. Since the Sn concentration was high in SAC305 alloy, when slight Sn diffused out of matrix to take shape into humps (Sn content was slightly higher in humps (100 wt. %) than that in the matrix (96.5 wt. %)), the surface tension decreased fleetly to form humps with different shapes, thus promoting the formation of Sn-rich patterns. As the spinodal decomposition processed, surface tension inside these Sn-rich humps was quite small and could not stimulate new humps to take shape. Consequently, no more structure changes took place after the spinodal decomposition. Rather, changes in microstructure due to spinodal decomposition were also observed in Sn-Pb, Co-Cu [
8] and Ag-Cu [
20] alloys, most of which were binary systems. Nevertheless, it is the first time that spinodal decomposition of the single component Sn in ternary Sn-Ag-Cu solder alloy was discovered, which is unprecedented that spinodal decomposition took place in solder alloys were all in binary systems.
Spherical solder powder is a new type of electronic solder material. It is widely applied as solder paste for electronic surface mount technology (SMT). The microstructure of the SAC305 solder powder also exhibited spinodal patterns when analyzed by SEM as
Figure 5 shows. The Sn-rich “worm” structures were about 2.5 µm in diameter. Additionally, the surface characteristics of SAC305 solder powder were basically the same as that of the solder ball as
Figure 2 presents. However, there were still some microstructure differences between the Sn-rich phase patterns on two products. On one hand, Sn-rich phase on solder powder ball surface closely connected each other with almost no trenches between them; on the other hand, Ag
3Sn precipitated more uniformly on the surface of solder powder than that on solder balls. It can be seen that many Ag
3Sn particles also distributed on the surface of Sn-rich phase, which was quite different from the phenomena that Ag
3Sn was mainly distributed in trenches between Sn-rich phase in solder balls. Such a discrepancy indicated that the precipitation sequence of Sn-rich phase and Ag
3Sn particles was not clear during the production process of solder powder. The starting composition fluctuations would undergo phase separation via spinodal decomposition, in which domains within the humps possessed local chemical enrichment of certain element Sn. A conclusion can be drawn by combining the results of BGA solder balls and powder that spinodal decomposition of Sn can happen very easily on spheroidal and ternary SAC305 solder alloy, which is supposedly owing to the reason that the interstitial diffusion in Sn can enhance the flux of thermal migration [
21]. Moreover, the oxygen in the protective gas during production of solder products also promoted the formation of Sn-rich phase since the driving force for the spinodal decomposition increased as the oxygen content increased [
22]. In experimental processes such as hot rolling, spinodal decomposition occurs at high temperatures. One of the most prominent features of solder alloys is that its microstructure is more prone to show phase coarsening and segregation characteristics, bringing about an increasing concern on reliability of SAC305 solder interconnects. In other words, the spinodal decomposed structure on surface of solder powder balls will affect the coating performance with the solder paste. In any case overall the material properties of the solder change over time which, eventually, has a detrimental effect on its joining capability.
The XPS patterns of Sn 3d peaks for surfaces of a SAC305 solder ball and powder samples are shown in
Figure 6. Both specimens showed similar features: the single peak curves of the surface were composed of three curves: Sn, Sn
2+ and Sn
4+ curves; there were two kinds of chemical states of Sn: Sn oxides and metallic Sn; the peak position of metallic Sn was located at about 484.8 eV, while the peaks standing for tin oxides Sn
2+ and Sn
4+ were showing at 486~487 eV. However, it can be seen that the peak of Sn
4+ in the ball was stronger than that in powder; on the contrary, another Sn
2+ peak was stronger in powder than that in the BGA ball. It turned out that there were various degrees of oxidative effect in both SAC305 solder balls and powder.
3.2. Microstructural Evolution by Phase Field Simulation
To the best of the authors’ knowledge, patterns of spinodal decomposition in high-temperature metallic liquids (particularly, in Sn-Ag-Cu ternary alloy) have not previously been observed and described in the available literature. Spinodal decomposition will not be observed in experiments because the transition is very fast. Therefore, results of phase field simulations were applied and also confirmed the above discovery about spinodal decomposition in the Sn-Ag-Cu ternary system to detect deeply the reaction process [
23]. Time evolution of the 2D microstructure version of phase separation was summarized in
Figure 7a–e. As can be seen at
t = 20 time in
Figure 7a, the microstructure was relatively fine and contained a large number of precipitates, leading to a small distance between phases which might cause an increase in the alloy strength because a higher opposition to the dislocation movement exists [
24]. There were more rod-like Sn-rich phases at the earlier stage of spinodal decomposition, which was presumed to be the reason that the composition at this moment had not fluctuated completely yet. As the aging time progressed, coarsening of the second phase through migration of the phase boundaries, dissolution, merging, and breakup could be easily inferred from the
Figure 7b. The large size Sn-rich phase particles grew up at the expense of the small size ones, thus forming the stochastic worm-like morphology as black dashed circle in
Figure 7d shows. The growth of the Sn-rich phase mainly occurred with Ostwald ripening process in which the small precipitates dissolved and absorbed by the larger ones, and as a result the number of precipitates became smaller with time. However, the number of the long-rod-shaped Sn-rich phase significantly reduced and rounder Sn-rich phase formed discretely when the reaction was almost complete at
t = 100. Conclusively, it was the incomplete spinodal decomposition that caused the formation of continuous rod-like Sn-rich phase as the black dashed circle in
Figure 7d shows; the discrete round Sn-rich phase region proved that the spinodal decomposition phase separation had reacted completely as the white dashed circle in
Figure 7e presents. It is worth indicating that the simulation results presented in
Figure 7 exhibit some discrepancies from the experimental observation, in which the rod-like Sn-rich phase particles distribute randomly rather than precipitate regularly with long sides connecting closely as shown in
Figure 2c. This is because the experimental observations are usually influenced by more factors, such as the formation of IMC, anisotropies of surface energy and grain boundaries [
9]. Despite these limitations, the simulated characteristics of the microstructure were largely consistent with the experimental observations. In conclusion, the formation mechanism of the Sn-rich phase pattern microstructure on the surface of SAC305 alloy is the spinodal decomposition of Sn according to results of both experimental investigation and phase field simulation.
Figure 8 shows the volume fraction of spinodal decomposition patterns and the evolution of total free energy according to reaction time in the phase field simulation results that show the kinetics of spinodal decomposition. The plot in
Figure 8a shows an increase in volume fraction with time up to about
t = 30. Then, the growth kinetic rate was decelerated, and subsequently the volume fraction remained almost constant. This fact indicated that the phase decomposition finished, and the coarsening of the decomposed phases began at about
t = 90. This stage was characterized by the growth of the Sn-rich phase at the expense of the smaller ones, as can be seen in
Figure 8b. It can be deduced that higher total free energy provided stronger fluctuation in concentration at early stage, which promoted the nucleation and subsequent growth of large and continuous rod-like Sn-rich phase. As phase separation proceeded, the large Sn-rich phase swallowed up small ones so as to form stable microstructure, which was consistent with the theory that large particles grew at the expense of smaller ones in diffusion-controlled coarsening [
25]. The decomposition was determined solely by diffusion of Sn element since there was no thermodynamic barrier to the reaction inside of the spinodal region [
7]. After this time
t = 120, the Sn-rich phase formed morphologically stable spherical precipitates as the interfacial energy of the precipitates reduced via coarsening, as expected from the Cahn-Hilliard model for spinodal decomposition [
11]. In conclusion, it is necessary to strictly control the aging time since this processing parameter does have a prominent effect on microstructure of solder alloys. Additionally, it is reported that the presence of the phase decomposition was responsible for the increase in hardness, which improved the overall property of the alloys [
26].
The thermal stress and temperature gradient at the interface between the separated phases and solder microstructure are high, which can promote the crack initiation and expansion and even cause deterioration of the reliability of solder joints [
27]. Thus, it is meaningful to characterize the microstructural evolution of such phase separation and coarsening of solders in interconnects. Nevertheless, the current model had some limitations. The phase field model only explained the corresponding spinodal decomposition phenomenon but did not correspond to the alloy composition of SAC305, which was included in our on-going investigation work. Furthermore, how to quantitatively control the area ratio of the Sn-rich phase and glossy part on the surface of solder balls was another meaningful task due to its effects on reliability of solder balls and powder, as mentioned in the introduction part of this article.