Effects of Isothermal Aging on Interfacial Microstructure and Shear Properties of Sn-4.5Sb-3.5Bi-0.1Ag Soldering with ENIG and ENEPIG Substrates

Abstract

Sn–Sb system solders and ENIG/ENEPIG surface finish layers are commonly used in electronic products. To illustrate the thermal reliability evaluation of such solder joints, we studied the interfacial microstructure and shear properties of Sn-4.5Sb-3.5Bi-0.1Ag/ENIG and Sn-4.5Sb-3.5Bi-0.1Ag/ENEPIG solder joints after aging at 150 °C for 250, 500 and 1000 h. The results show that the intermetallic compound of Sn-4.5Sb-3.5Bi-0.1Ag/ENIG interface was more continuous and uniform compared with that of Sn-4.5Sb-3.5Bi-0.1Ag/ENEPIG interface after reflow. The thickness of the interfacial intermetallic compounds of the former was significantly thinner than that of the latter before and after aging. With extension of aging time, the former interface was stable, while obvious voids appeared at the interface of the latter after 500 h aging and significant fracture occurred after 1000 h aging. The shear tests proved that shear strength of solder joints decreased with increasing aging time. For the Sn-4.5Sb-3.5Bi-0.1Ag/ENEPIG joint after 1000 h aging, the fracture mode is ductile-brittle mixed type, which means fracture could occur at the solder matrix or the solder/IMC interface. For other samples of these two types of joints, ductile fracture occurred inside of the solder. The Sn-4.5Sb-3.5Bi-0.1Ag/ENIG solder joint was thermally more reliable than Sn-4.5Sb-3.5Bi-0.1Ag/ENEPIG.

1. Introduction

Sn–Pb solder, as a connection material for electronic products, has always been favored by the electronics industry due to its excellent technological performance and low cost. However, because of the toxicity of Pb, countries in succession set off a wave of “lead-free” after entering the 21st century. At present, electronic manufacturing is in a special stage of transition from lead-containing to lead-free soldering. Among the several lead-free solders, the ternary alloy Sn–Ag–Cu is regarded as the leading candidate system for reflow soldering. However, the lower melt point of the alloy limits its application in high temperature. The Sn–Sb system solders perform better in terms of cost and high temperature applicability [1].

Due to the solid solution strengthening of Sb element and the formation of Sn–Sb intermetallic compound (IMC), Sn–Sb solders have good creep resistance and higher strength [2,3,4,5]. In order to further improve performance, ternary or quaternary alloys are developed by adding alloying elements. For example, the pronounced grain refinement is achieved as the addition of Ag [6]. Adding element of Bi led to solid solution strengthening and enhanced thermal reliability of Sn–Sb alloy [7].

Study of interfacial reactions between solders and metallization layers is a crucial subject for the reliability evaluation of the solder joints. Studies have shown that the stability of the solder joint interface largely depends on the IMC generated by the metallurgical reaction between the solder and the substrate surface during the soldering process [8,9,10]. After thermal aging, the thickness of the intermetallic compound will increase, and its own brittleness and the Kirkendall voids caused by the unequal diffusion of atoms during the growth process will severely weaken the mechanical properties of the solder joints. Therefore, it is very necessary to study the interface microstructure and mechanical properties of solder joints after isothermal aging. Cu is commonly used as a metallization layer in the soldering process. As Cu substrate contacts with solder directly during high temperature reflow soldering, the Cu element quickly dissolve and diffuse into the solder, which is prone to overreaction and affects the reliability of solder joints [11,12,13,14]. Thus surface treatments are applied to improve the interfacial properties of solder joints. Among many surface treatments, electroless nickel-immersion gold (ENIG) is the most mature one [15,16]. A layer of Ni is coated on the Cu substrate, and then a layer of Au is coated on the Ni layer, where Ni and Au coatings act as the barrier and anti-oxidation layer, respectively. Recently, the interfacial reaction between Sn—A–-Cu solder and ENIG surface finish has been widely studied. The results indicated that (Ni, Cu)₃Sn₄ phase formed at ENIG/Sn-3.5Ag-0.7Cu/ENIG sandwiched interfaces after reflowing. After aging at 150 °C for 24 h, the IMC became (Cu, Ni)₆Sn₅ phase. The interface composition did not further change after aging for 150 h. The IMC changed back to (Ni, Cu)₃Sn₄ phase after aging for 500 h [17]. The (Ni, Cu)₃Sn₄ and (Cu, Ni)₆Sn₅ interface compound were also found at Sn-4.0 wt % Ag-0.5 wt % Cu/ENIG interface, and voids were found inside the Sn–Ni–Cu ternary interfacial compounds after 500 and 1000 h aging [18].

However, due to the complexity of ENIG process, the phenomenon of “black pad” will appear in the process of actual use, which affects the reliability of soldering [19]. In addition to ENIG, electroless nickel-electroless palladium-immersion gold (ENEPIG) as a feasible surface finish has been widely used in various fields [20]. The Pd layer used as a protective layer to prevent the corrosion of the nickel (phosphorus) layer during the gold plating process and effectively solve the black pad problems [21]. Simultaneously, the interfacial reactions and mechanical properties of Solders/ENEPIG are extensively investigated. Chien-Fu et al. [22] studied the interfacial reaction and mechanical strength of Sn-3.0 wt % Ag-0.5 wt % Cu jointed with ENEPIG surface finish. The needle-like (Cu, Ni, Pd)₆Sn₅ formed in the ENEPIG joints. By inspecting the fracture, the ductile fracture was observed in the ENEPIG joint.

Nevertheless, data and reports on comparative studies between ENIG and ENEPIG for the interfacial reactions are still seriously lacking, especially combined with high-temperature Sn–Sb solder. In this work, the effects of isothermal aging on interfacial microstructure and shear properties of Sn-4.5Sb-3.5Bi-0.1Ag jointed with ENIG and ENEPIG surface finishes were investigated. The results would provide theoretical basis for evaluating the reliability of two solder joints.

2. Materials and Methods

In order to prepare the ENIG and ENEPIG substrates, the Au/Ni bilayer with the thickness of 0.1 μm/4 μm and the Au/Pd/Ni trilayer with the thickness of 0.1 μm/0.2 μm/4 μm were deposited on the Cu pads. The component is chip ceramic capacitor 0402, and the substrate is a 28 mm × 28 mm, 1.6 mm thick multi-layered (ground copper planes) FR-4 printed circuit board. The Sn-4.5Sb-3.5Bi-0.1Ag solder was printed on the ENIG or ENEPIG substrate by using an automatic solder paste and reflowed at a peak temperature of 280 °C for a belt speed of 1m/min in a vacuum reflow machine (Vacm-0023, Germany) to complete the surface mounting process. Here, the prepared Sn-4.5Sb-3.5Bi-0.1Ag/ENIG and Sn-4.5Sb-3.5Bi-0.1Ag/ENEPIG joints are named S1 and S2 respectively. The temperature profile of the reflow treatment and morphology of S1 after soldering was shown in Figure 1. In order to study their thermal aging properties, the two kinds of samples were aged in a high temperature test chamber (HT305E) at 150 °C for 250, 500 and 1000 h respectively.

The interfacial morphology of the joints was observed by scanning electron microscope (SEM; ZEISS ULTRA55, Carl Zeiss, Jena, Germany) after spraying gold on the surface of the specimens. To reveal the cross-sectional microstructure, the specimens were prepared by grounding, polishing and etching using 93%C₂H₅OH-5%HNO₃-2%HCl etchant. Energy dispersive X-ray spectroscopy (EDS; Carl Zeiss, Jena, Germany) and micro-region X-ray diffraction (XRD; Bruker, Spabrucken, Germany) analysis were performed to identify the phases. The thickness of the IMC was carried out using image processing software (Adobe Photoshop CS6, Adobe, San Jose, America). The area pixel value A and length pixel value L of the intermetallic layer are measured by using Photoshop software, the formula $\mathrm{h}=\mathrm{A}/\mathrm{L}$ to find the number of pixel in the thickness of the intermetallic layer, and then obtain the actual length represented by each pixel according to the ratio of the scale value to the scale pixel value. Finally, the actual thickness of the interface IMC layer is obtained. The shear strength of solder joint was tested by MFM series bond tester, the shear height is 50 μm, and the shear speed is 700 μm/s. After shear testing, fracture and cross section morphologies were observed and analyzed by SEM and EDS.

3. Results and Discussion

3.1. Original Morphology and Compositions of IMC

Figure 2 shows cross-sectional SEM images and EDS results of the interfaces of S1 and S2. The EDS point analysis indicated that Ni₃Sn₄ layer formed at the interface of S1, which was continuous and uniform, showing a typical scallop shape, as presented in Figure 2a,b. At the interface of S2, a discontinuous but protruding massive IMC is formed. The EDS analysis indicated the massive IMC was in the (Pd, Ni)Sn₄ phase, whose average atomic fractions are 78.14 at.%, Sn-4.38 at.%, and Ni-17.48 at.% Pd, as illustrated in Figure 2c,d. In addition, the detachment of the (Pd, Ni)Sn₄ phase away from the Ni layer was found in Figure 2c.

For the two solder joints, no Au compounds were detected in the interface. Since the thickness of the Au coating layer is approximate 0.1 μm. At the beginning of the reflow, the Au layer quickly dissolves into the solder within a few seconds. Thus, the Ni layer of ENIG coating exposed to the solder side and reacted with molten solder to form the Ni₃Sn₄ phase. After Au atoms consumed out at the interface, Pd layer of the ENEPIG coating is sequentially exposed to the solder. Unlike Au surface layer, the thickness of the intermediate Pd layer is about twice as much as the Au layer, and the dissolution rate of Pd in the solder is about two orders of magnitude lower than the dissolution rate of Au [23]. The Pd layer reacted with Sn to form PdSn₄ phase, as reported by Tanaka S et al. [24] There are extensive interaction between Pd and Ni, and the Ni atoms diffuse to the Pd-containing IMCs with a large flux. We speculate that as Pd atoms leaves the interface, Ni atoms can dissolve into the PdSn₄ phase and replace the position of Pd atom to form (Pd, Ni)Sn₄ phase.

3.2. The Impact of Aging Treatment on IMC

Figure 3 is cross-sectional SEM micrographs of the solder joints aged at 150 °C, where the left side shows interfacial microstructure of S1 after thermal aging. When the samples were subjected to thermal aging, they exhibited different microstructural evolution at the interfaces. The scallop shape morphology of the interface is no longer so obvious, and the entire interface tends to be flat after 250 and 500 h aging, illustrated by Figure 3a,b. Figure 4a shows the XRD pattern of the interface of S1 after aging for 250 h, the interface compound is still Ni₃Sn₄. When the aging time was extended to 1000 h, a new phase was formed at the interface, as seen by the arrows in Figure 3c. The EDS analysis indicated the average composition of this new phase was 7.34 at.% Cu-35.77 at.% Ni-56.89 at.% Sn, as presented in Figure 5, which was identified as the (Ni, Cu)₃Sn₄ phase. To explain how the Ni₃Sn₄ phase changed into (Ni, Cu)₃Sn₄ phase, we conducted series of EDS mapping analysis. Figure 6 is EDS mapping result of the interface of S1 after aging for 1000 h at 150 °C. It can be seen that the Ni layer becomes discontinuous, resulting in several gaps. We speculate that these gaps provide channels for diffusion of Cu atoms in the substrate and Cu can interact with Ni atoms at the interface. As Cu had similar crystal structure with Ni, Ni atoms were gradually replaced by the Cu atoms to form the (Ni, Cu)₃Sn₄ phase. The aging test proved that the interface of S1 is stable, and no voids or cracks were found.

The right side of Figure 3 shows the morphology of the interface of S2 after aging. Figure 4b is XRD pattern of the interface of S2 aged at 150 °C for 250 h, which indicates that Ni₃Sn₄ phase formed. Related research found that the (Pd, Ni)Sn₄ phase will spall into molten solder after reflow, and as the reaction progresses, the spalled (Pd, Ni)Sn₄ phase will dissolve into the molten solder [25]. As described earlier, the spalled massive (Pd, Ni)Sn₄ phase was observed at the interface of S2, as seen in Figure 2c. But after aging, (Pd, Ni)Sn₄ phase had completely dissolved into the solder. Simultaneously, the Ni layer exposed and Ni₃Sn₄ phase formed. Kirkendall voids are generated in IMC when aging at 150 °C for 500 h. The voids are so tiny that will only be discovered after long time aging. The cause of Kirkendall voids is due to the difference in the diffusion rate between atoms. This unbalanced diffusion mechanism leads to the generation of atomic-level holes, which is accompanied by the Kirkendall effect during the reflow process. The aging experiment accelerated the occurrence of the Kirkendall effect and prompted the Kirkendall voids to gather and grow, forming larger voids which are easy to be observed. A continuous fracture dramatically occurred along the interface after aging for 1000 h. The thermal stress is suggested to be affected by the cooling process, and during which the fracture initiated from the local voids and propagated across the entire interface.

The Figure 3d,h show the morphology of the Ni₃P crystallization layer of S1 and S2 after aging 1000 h, respectively. The Ni₃P crystallization layer is stable and no voids were found. According to the study of Tseng et al. [26], on the case of interfacial reaction of Sn–3.0Ag–0.5Cu solder jointed with ENIG and ENEPIG surface finishes, Ni₃Sn₄ was inhibited in the ENEPIG joints, which suppressed the excessive voids formation as compared to ENIG during thermal aging. In this case, Cu₆Sn₅ phase would form in the interfacial reaction of Sn–3.0Ag–0.5Cu solder with electroless Ni–P/immersion Au (ENIG) and electroless Ni–P/electroless Pd/immersion Au (ENEPIG). Moreover, Ni₃Sn₄ phase could occur in the ENIG aged joints, but not in the ENEPIG aged joints. It was demonstrated that the Ni₃Sn₄ phase usually accelerate the growth of columnar Kirkendall voids inside the Ni₃P layer, thus resulting in poor performance of welded joints. However, the Sn–Sb–Ag solder used in this paper does not contain the Cu element, the Cu₆Sn₅ phase would not form at the interface of S1 and S2 and only the Ni₃Sn₄ phase formed in both aged joints. In addition to Ni₃Sn₄ phase, it is proved that the ratio of phosphorus to nickel in the Ni coating layer and the soldering procedure will also affect the formation of Ni₃P layer. For Sn–Sb–Ag solder, we adopted reasonable soldering process that make Ni₃P crystallization layer without columnar voids.

The IMC thickness of the solder joints increases with increasing aging time. It is generally believed that the rate controlling mechanism for the growth of the IMC is a diffusion process [27]. Figure 7 shows a function of the IMC thickness and the square root of the aging time. The IMC layer thickness increased linearly with the square root of aging time, which can be described by the following formula: $X-X_0=\sqrt{Dt}$, where the symbols $X$, $X_0$, $D$ and $t$ represent the thickness of IMC layer after and before growth, growth rate constant and aging time, respectively. By calculation, the growth rate constants D of S1 and S2 interfacial compounds are 1.298 × 10⁻¹⁴ and 5.516 × 10⁻¹⁴ cm²/s, respectively. With increasing aging time, Ni atoms in ENIG and ENEPIG continuously diffuse into the solder, resulting in growth of the interface. Since the presence of Pd can increase the growth of IMC [28], thus the IMC thickness of S2 is thicker than that of S1.

3.3. Mechanical Properties

The relationship between shear strength and aging time is shown in Figure 8. The shear strength of the solder joint tends to be stable. After 1000 h aging, the shear strength of the S1 and S2 decreased by 5.48% and 10.85%, respectively. S2 exhibits lower shear strength than S1 during the aging process.

As we know, shear failure can occur in solder, solder/IMC interface or IMC, corresponding to ductile fracture, ductile-brittle mixed fracture and brittle fracture modes, respectively. Figure 9 shows the fracture morphologies of solder joints before and after different time aging at 150 °C. Figure 9a, f are the macroscopic fracture morphology of S1 and S2, respectively. The macroscopic morphology of the fracture is mainly composed of two parts. The fracture morphology at position 1 is basically the same, so we mainly analyzed the microscopic morphology of the fracture at position 2. The left-hand side columns show the fracture morphologies of S1. There are densely and small-sized dimple structures distributed at the fracture before aging, as indicated in Figure 9b. The dimples of the shear fracture became larger and the number decreased after aging for 250 h, as shown in Figure 9c. The dimples become lesser and local smooth areas appeared after aging for 500 h. And after 1000 h of aging, the local smooth areas have increased with increasing aging time.

Figure 10 shows the cross-sectional view of fracture surfaces of solder joints after different aging time under 150 °C. The left-hand side columns show the interfaces of S1. As presented in Figure 10a–d, the fracture position of the solder joint occurred in the solder matrix and gradually approaches the IMC interface. This trend was obvious in most tested samples. For each kind of solder joints aging at different time, five samples were analyzed. Due to the influence of artificial experimental operations, the five samples of each combination of solder alloy have different fracture positions in the solder matrix, but with the increase of aging time, the fracture position of the samples show a tendency to gradually approach the IMC layer. Related studies indicate that for the aged samples, as the aging time increases, the shear fracture of the solder joint usually occurs in the IMC layer [29]. However, Figure 10d shows that the shear fracture of the S1 still occurred in the solder after aging for 1000 h. On the one hand, IMC thickness of this solder joint is relatively thin, and the morphology is flat, resulting in a more uniform stress distribution. On the other hand, Cu–Ni–Sn IMCs have high shear modulus [30]. Therefore, the shear forces tears the solder matrix of S1 into elongated dimples, and the failure mode is still ductile fracture after long time aging.

The right-hand side columns show the fracture morphologies of S2 before and after different time aging at 150 °C. Same to the fracture morphology of S1, there is densely and small-sized dimple structures distributed at the fracture surface of S2 before aging, as indicated in Figure 9g. However, local smooth areas appeared after aging for 250 h and there are nearly no dimples on the fracture surface after 500 h aging. Moreover, it can be found from Figure 9j that pits were formed on the fracture surface of the solder joint which was aged for 1000 h, and the bottom of pits were composed of Ni₃Sn₄ phase. Combining with Figure 10h, it can conclude that one part of the fracture occurred at the interface, and the other part occurred in the solder matrix. The fracture mode of the solder joint is ductile-brittle mixed, with the lowest strength. The IMC thickness of S2 is higher than S1 before and after aging, and the IMC shape of the former is irregular. The thicker IMC layer and irregular shape of the IMC increase the roughness of the S2 interface, which causes large local stress of the interface under the shear force and partial fracture on the solder/IMC interface.

4. Conclusions

In this study, a comparative analysis of interfacial microstructure and shear properties of Sn-4.5Sb-3.5Bi-0.1Ag solder with two kinds of surface finish layers (ENIG and ENEPIG) was conducted. The conclusions are as follows:

(1) For the Sn-4.5Sb-3.5Bi-0.1Ag/ENIG interface, a continuous and uniform scallop-like Ni₃Sn₄ interface compound is formed after reflow. With the increase of aging time, the thickness of the interface IMC layer gradually increased, and no obvious voids and fracture generated at interface. When the aging time was 1000 h, the initial thickness increased from 1.86 μm to 4.05 μm, and the interface is consisted of (Ni, Cu)₃Sn₄ phase.

(2) For the Sn-4.5Sb-3.5Bi-0.1Ag/ENEPIG interface, a massive discontinuous (Pd, Ni)Sn₄ phase formed after reflow. After aging, the IMC consisted of the Ni₃Sn₄ phase. With the increasing of aging time, the interface thickness increases continuously and Kirkendall voids are generated. After aging for 1000 h, a continuous fracture occurred.

(3) The shear strength of Sn-4.5Sb-3.5Bi-0.1Ag/ENIG and Sn-4.5Sb-3.5Bi-0.1Ag/ENEPIG solder joints decrease with the increase of thermal aging time. After 1000 h aging, the shear strength of the two solder joints decreased by 5.48% and 10.85%, respectively. The fracture mode of the Sn-4.5Sb-3.5Bi-0.1Ag/ENIG solder joint is ductile after reflow and aging. The fracture type of Sn-4.5Sb-3.5Bi-0.1Ag/ENEPIG solder joint is ductile after reflow and 250, 500 h aging. When the aging time was extended to 1000 h, the fracture pattern transformed into ductile-brittle mixed mode.

As stated previously, the Sn-4.5Sb-3.5Bi-0.1Ag/ENIG solder joint is thermally more reliable than Sn-4.5Sb-3.5Bi-0.1Ag/ENEPIG.