Usually, the IMC with thickness of 1–3 μm in solder joints is considered to process sufficient reaction between copper pillar bump and metallization pad. When thick IMC forms, voids and crack may occur in copper pillar joints, leading to shorter service time. Thus, it is necessary to study the growth rule of IMC in solder joints. The temperature and time are the two key parameters to control the IMC growth rate, with constant load of 1 MPa. The reaction temperature sets at 230°C, 240°C, and 250°C, respectively, and the reaction time is from 1 min to 10 min. The IMC thickness was calculated by the IMC area divided by the IMC length, while the IMC area and length were tested from SEM image.

Figure 2 indicates the relationship between heating parameters and IMC thickness of solder joints. Mathematical equations were used to fit the experimental data. At different heating temperature, the IMC is growing gradually with time increases, basically according to the equation of d = D t where d is the thickness of IMC, t is the time, and D is the growth rate. By calculation, the growth rates of IMC are 0.09 μm/min at 230°C, 0.14 μm/min at 240°C, and 0.19 μm/min at 250°C. An obvious rule can be obtained that higher temperature can accelerate the growth rate of IMC. The reason is that higher temperature can increase the solution rate of Ni and Cu atoms to liquid solder, while the higher temperature beneficially accelerates the atomic reaction rate to form IMC. Longer time will extend the reaction to form more IMC at interface.

Figure 3 illustratesthe microstructure of IMC at different temperature and time. When sample sinter at 230°C for 1 min, very thin IMC formed and the elementary compositions are (Ni,Cu)₃Sn₄ and (Ni,Cu)₃Sn. The result is similar to previous studies (Dai et al., 2022). When reaction time extends to 10 min, besides obvious thickness increase, the elementary composition transforms into (Cu,Ni)₆Sn₅. No more phase transformation is observed indicates that the (Cu,Ni)₆Sn₅ is stable crystal structure. When heating time increases to 240°C, the initial crystal structure of IMC is (Ni,Cu)Sn₃, and it will also gradually transform into (Ni,Cu)₆Sn₅ at 10 min. By further increasing temperature to 250°C, the microstructure of IMC transforms quickly, and the initial (Ni,Cu)Sn₃ will become to (Ni,Cu)₆Sn₅ at 5 min. Based on these research, it can be seen that the (Ni,Cu)₃Sn₄, (Ni,Cu)₃Sn, and (Ni,Cu)Sn₃ are transition crystal structures, and those structures will be eventually transformed into stable (Ni, Cu)₆Sn₅ structure. The reason for the crystal structure transformation principle is that, at the beginning of the soldering reaction, the liquid solder would provide sufficient Sn atoms. Thus, the Sn-rich crystal structures primarily grow. However, the formation energy of the alloy (Ni,Cu)₆Sn₅ is negative, so it is more thermodynamically stable and its alloying capacity is higher, resulting in automatic transformation from transition crystal structures to (Ni,Cu)₆Sn₅ structure (Leineweber et al., 2021). In summary, when there is sufficient atomic diffusion time, the final crystals structure is the (Ni,Cu)₆Sn₅ structure. Meanwhile, high temperature would accelerate the atomic diffusion rate, leading the faster crystal structure transformation from transition crystal structures to the most stable (Ni,Cu)₆Sn₅ structure.

To study the effects of IMC on the electromigration lifetime of solder joints, a daisy chain of copper pillar bumps was designed. The daisy chain is composed of two substrates as shown in Figure 4. The copper pillar substrate contains two lead bonding pads, which are used to connect DC power. A 6 × 6 copper pillar array layouts on the substrate, and a metal wire prepared by electroplating connects each two copper pillars, with all copper pillars at non-conducting states. The other pad substrate has a 6 × 6 metallization pad array. Each two pads are connected by metal wire, with all pads at non-conducting states. The two paired substrates were bonding by hot pressing methods.

Based on the Arrhenius equation, improving temperature can significantly accelerate the atomic diffusion rate. Thus, to accelerate the atomic diffusion and shorten the research period, all experiments were conducted at environment of 100°C. The current density through solder joints sets at 1.0 × 10⁴ A/cm². The experimental samples were placed on a hot plate, and the heating rate was 5°C/min. When the temperature reached at set value, the current stress was immediately loading to the samples. The failure criterion of solder joints is electrical open. Three kinds of solder joints were prepared to study the effects of IMC on lifetime of solder joints. Table 1 shows the IMC thickness and resistance of copper pillar daisy chain samples. The IMC thickness of thin, thick, and all IMC samples are 1.3 μm, 2.7 μm, and 10.1 μm, respectively. Figure 5 shows the initial microstructure of solder joints after reflow.

Figure 6 shows the resistance change with time increase and cross-sectional image of failed solder joints. At simultaneous stresses of heat and electron flow, the resistance evolution of thin IMC sample could be divided into three steps. In the first step, the resistance has a huge increase from 2.3 Ω to 4.9 Ω. The obvious resistance change is due to quickly IMC increase. As we know, the conductivity of IMC crystal structure is obviously lower than that of Sn96.5Ag0.5 solder (The electrical resistivity of Cu₆Sn₅ and Sn is 17.5 mΩ cm and 10.9 mΩ cm, respectively) (Dai et al., 2022)^, (Zhu et al., 2020). Thicker IMC will continuously reduce the electric conduction of joints. In the second step, resistance of joint increases to 5.1 Ω and trend to be stable until to 250 min. After that, the resistance increases significantly until to 400 min when the daisy chain electrically failed. By observing the failed sample (Figure 6B), a crack close to the cathode side crosses the solder joint, but at the right corner of solder joints destruction is more serious. When the electron flow transports from cathode to anode, quick crystal structure transformation generates lots of voids in copper pillar and solder interface, which will finally aggregate to crack. Additionally, because the current tends to transmit along the least resistance pathway, there is a current crowding in the right corner. Therefore, the right corner near the cathode of solder joint has the highest failure risk.

The same method is applied to monitor the resistance evolution of the thick IMC sample. Indicating in Figure 7A, there is a sharp increase in the initial few minutes, in which the resistance increases from 2.8 Ω to 5.1 Ω. In the second step after 10 min, the state of resistance is stable firstly and has a slight increasement after 150 min. Then, the growth rate of resistance gradually increases until to failure. Figure 7B shows microstructure of failure sample, lots of cracks form in the middle area of the joint. In this sample, initial IMC layer distance between copper pillar side and metallization pad side is less than 5 μm, voids generate at both anode and cathode will aggregate together. Thus, the time for crack crossing the solder joint unexpected shorten, resulting in a lifetime of thick IMC sample.

Unlike the thin and thick IMC samples, there is no quick resistance increase step being observed in all IMC joints, as shown in Figure 8A. Because all IMC crystal structure has formed completely. The lifetime of all IMC sample can reach about 1,200 min Due to the fact that there is no quick phase transformation during aging process, the failure of full IMC joint is due to high Joule heat caused by higher resistivity of IMC structure. In addition, the coefficient of thermal expansion (CTE) mismatch between pad materials and IMC cause mechanical stress in joints, which leads to crack initiation near cathode, as shown in Figure 8B. In the middle of joint, EM leads to several voids, but there are much less voids generating compared to thin and thick IMC joint. Under the electrical stress, the atoms of joints are moved by the electron wind force, leaving lots of voids in the joints, which is the main reason for EM failure. Because the interatomic force in IMC is much stronger than that in Sn96.5Ag3.5 solder, the atomic movement rate in all IMC joints is much slow than that in Sn96.5Ag3.5 solder. In summary, EM is not major factor of joint failure due to IMC excellent resistance to EM.

By comparison, the solder joints with nearly 100% percent of IMC crystal structure have the highest lifetime, which is triple lifetime of thin IMC sample and four times of lifetime of thick IMC sample. The main reason is that all IMC sample can greatly reduce the number of voids, which are generated by crystal structure transforming. The research results in this study provide a novel solution to design long lifetime solder joints.