1 Introduction

In recent decades, electronic power modules have undergone rapid evolution, integrating higher input/output density, increased chip counts, and enhanced heat dissipation capabilities [1, 2]. With the advent of the post-Moore’s Law era and the emergence of new electronic product domains such as the Internet of Things (IoT), 5G, high-capacity storage, high-performance computing, autonomous driving, and holographic technology, electronic components are increasingly moving towards miniaturization and multifunctionality [3, 4]. This trend towards miniaturization and diversification has sparked widespread interest in three-dimensional integrated circuit (3D-IC) technology, where microbumps play a crucial role [5]. As electronic products have become smaller, the size of microbumps has also been reduced to less than 100 µm. However, this significant size reduction often results in changes to the mechanical properties and microstructures of microbumps, leading to mechanical reliability issues [6]. As the volume of solder in microbumps decreases, the proportion of Intermetallic Compounds (IMCs) increases after the reflow process, which significantly impacts the reliability of electronic products. Interface reactions induced by IMC formation, such as volume shrinkage or grain collisions, become more pronounced at the microbump scale.

Ni/Sn/Ni is one of the most commonly used material systems in microbumps. In other material systems, multiple IMCs layers would form, which become weak points [7]. Due to the superior thermal stability of Ni–Sn IMC, it does not undergo additional phase transformations during multiple reflow processes, and the only IMC at the Ni/Sn/Ni interface is Ni3Sn4[8], contributing to the higher shear strength of its joints compared to other material systems[7]. However, a critical issue arises in Ni/Sn/Ni joints within confined spaces. When Ni reacts with Sn, the volume shrinkage caused by the formation of Ni3Sn4 leads to the formation of voids along the centerline of the joints. These voids weaken the solder joint, raising concerns about mechanical reliability.

Early studies showed that in traditional Ball Grid Array (BGA) solder joints, Ag3Sn typically exhibits a plate-like structure, forming continuous surfaces that could lead to easy crack propagation [9, 10]. However, in micro-BGA, the plate-like Ag3Sn can acted as obstacles against shear force and enhanced the shear strength [11]. Besides, when Ag3Sn was present as nanoscale particles, it enhanced the tensile strength and Vickers hardness of the solder joint [12]. Ag3Sn exhibits varying effects on mechanical properties across different scales. Recent findings noted that adding Ag to Ni/Sn solder results in the formation of Ag3Sn, which helps fill the voids in the joints, the presence of Ag3Sn at the microbump scale seems to prevent voids from becoming stress concentration points [13, 14]. Additionally, studies investigating the microstructure evolution of Ni/Sn-xAg/Ni sandwich structures with different silver contents (x = 2.4%, 3.5%, 8.0% wt.%Ag) found that the optimal silver addition is 3.5 wt%, leading to the formation of bulk-type Ag3Sn to fill the voids due to volume shrinkage [15]. However, a complete explanation of the mechanical reliability of this system remains unresolved. This study aims to investigate the impact of adding Ag in the solder on mechanical performance and to realize the microstructure of Cu/Ni/Sn/Ni/Cu and Cu/Ni/Sn-3.5Ag/Ni/Cu with a bond thickness below 10 µm.

2 Experimental methods

In this study, Cu substrate with an electroplated Ni layer was used substrates in a sandwich structure to prepare microbumps. Four experimental conditions of sample were fabricated, including Cu/Ni/Sn/Ni/Cu (200 s), Cu/Ni/Sn-3.5Ag/Ni/Cu (200 s), Cu/Ni/Sn/Ni/Cu (400 s), and Cu/Ni/Sn-3.5Ag/Ni/Cu (400 s), shown in Fig. 1d. To achieve better wettability between electroplated Ni layer and solder, the solder used in this experiment will be Sn-3.5Ag foil. Firstly, Sn-3.5Ag solder balls with a diameter of 300 µm were placed between glass slides and heated. The surface of Sn-3.5Ag solder balls was cleaned from oxides using water-soluble flux. When heated above the eutectic point, the Sn-3.5Ag solder balls melted to form soldering foil, as shown in Fig. 1a, b. Ag element in the Sn-3.5Ag foil remains uniformly distributed after heating process, as shown in Fig. 1c. The Cu/Ni structure was prepared by electroplating a 5 µm thickness Ni layer on the Cu substrate. The upper substrate was cut to a size of 2 mm*2 mm*0.4 mm, and the lower substrate was cut to a size of 2 mm*10 mm*0.4 mm. All substrates were properly ground and cleaned by immersing in 1.5 M HCl. Then, water-soluble flux was applied to the material surface to remove oxides and contaminations. The reflow process was conducted at a peak temperature of 260 °C for 200 s and 400 s, while a fixed pressure of 1.6 MPa was simultaneously applied to the samples. Finally, the joints were air-cooled to room temperature. The bump height of the solder joints was maintained at 8 ± 1 μm.

Fig. 1
figure 1

a Preparing soldering foil process b Bonding process of samples in this work, c Distribution of Ag element in the Sn-3.5Ag foil, d Schematics of test design: Cu/Ni//Sn/Ni/Cu, Cu/Ni//Sn-3.5Ag/Ni/Cu sandwich structure

Full size image

To obtain microstructural information of the samples, samples were cold-mounted by epoxy resin after the reflow process, and were ground with sandpapers. Further polishing was carried out using a cross-section polisher (CP) for high-quality surfaces. Field emission scanning electron microscopy (FE-SEM; JSM-7600F, JEOL) was used for cross-sectional observation of samples. A field emission electron probe micro-analyzer (FE-EPMA; iHP-F200, JEOL) was used to obtain the composition of different phases for phase identification. Electron back-scattered diffraction (EBSD) was utilized to gain information about grain sizes and grain orientations. Shear strength was evaluated by a bond tester (Condor Sigma, Xyztec). The shear speed was 100 μm/s, and the shear height was 40 μm.

3 Results and discussion

Figure 2a–d are the cross-section back-scattered electron images corresponding to Cu/Ni/Sn/Ni/Cu (200 s), Cu/Ni/Sn/Ni/Cu (400 s), Cu/Ni/Sn-3.5Ag/Ni/Cu (200 s), and Cu/Ni/Sn-3.5Ag/Ni/Cu (400 s), respectively. When the reflow time is extended to 400 s, the thickness of (Ni, Cu)3Sn4 increases. As Ni reacts with Sn to form (Ni, Cu)3Sn4, and voids would form along the centerline of the joint due to volume shrinkage, as shown in Fig. 2b. Since Sn, as the primary diffusing species, diffuses more extensively than nickel during the reaction process, voids appear at the original sites of the residual Sn [13, 16]. However, in samples with a reflow time of 200 s, no voids are observed because there is still sufficient Sn in the joint to compensate for the volume shrinkage, shown in Fig. 2a. After 200 s of reflow, the residual Sn matrix remains present in Cu/Ni/Sn-3.5Ag/Ni/Cu, as shown in Fig. 2c. As the reflow time increases, grains of (Ni, Cu)₃Sn₄ grow from opposite directions and begin to impinge on each other. When the residual Sn is almost consumed, (Ni, Cu)₃Sn₄ nearly occupies the Ni/Sn-3.5Ag/Ni interface, and bulk-type Ag3Sn were observed, instead of void formation [14], as shown in Fig. 2d.

Fig. 2
figure 2

Back-scattered electron images of a Cu/Ni/Sn/Ni/Cu (200 s seconds) b Cu/Ni/Sn/Ni/Cu (400 s) c Cu/Ni/Sn-3.5Ag/Ni/Cu (200 s) d Cu/Ni/Sn-3.5Ag/Ni/Cu (400 s)

Full size image

Quantitative analysis of Cu/Ni/Sn-3.5Ag/Ni/Cu samples after 400 s reflow by FE-EPMA were utilized to verify the composition of IMCs as listed Table 1. The compositional analysis points 1–3 were as (Ni, Cu)3Sn4, indicating that Cu atoms from the substrate diffused into the solder matrix through the columnar grain boundaries of the electroplated layer after reflow process [17]. Cu atoms uniformly distribute in the microbump and replace lattice positions of Ni, resulting in the formation of (Ni, Cu)3Sn4 instead of Ni3Sn4, as shown in Fig. 3d.

Table 1 Elemental quantification of IMCs in Cu/Ni/Sn-3.5Ag/Ni/Cu samples by FE-EPMA
Full size table
Fig. 3
figure 3

FE-EPMA X-ray mapping images of Cu/Ni/Sn3.5Ag/Ni/Cu (400 s) a back-scattered electron image, b Sn element, c Ag element, d Cu element

Full size image

To more clearly explain the IMC distribution, the elemental color mapping of Cu/Ni/Sn-3.5Ag/Ni/Cu (400 s) is revealed in Fig. 3. Through the Ag element mapping shown in Fig. 3c, two morphologies of Ag3Sn IMC, finely dispersed Ag3Sn particles and bulk-type Ag3Sn, can be observed after the reflow process, and bulk-type Ag3Sn tends to aggregate near the centerline to fill the voids [18]. The mechanisms for the formation of the two morphologies of Ag3Sn IMCs are believed to be different. In the reflow process, firstly, as Sn reacts with Ni to form (Ni, Cu)3Sn4, the amount of Sn matrix in the solder decreases. Ag does not react with Ni but only reacts with Sn to form Ag3Sn as the sole product at the Ni/Sn-3.5Ag/Ni joints [15, 19]. Due to the different growth rates of (Ni, Cu)3Sn4 in the solder, it is expected that the Ag concentration in certain localized regions will increase with increasing reflow time [20]. Secondly, due to the increasing Ag concentration in specific regions, the segregation of primary Ag3Sn becomes easier during the reflow process, shown in Fig. 4b. As primary Ag3Sn continues to coarsen, smaller grains tend to dissolve first, which is also known as Oswald ripening principle [21]. These solutes then diffuse through the liquid solder, and redeposited near larger grains. In this case, smaller Ag3Sn particles dissolve into the Sn-3.5Ag liquid solder to release Ag atoms, which diffuse to the vicinity of primary Ag3Sn and reprecipitate during the reflow process, promoting further growth of primary Ag3Sn to form bulk-type Ag3Sn [22], shown in Fig. 4c. The driving force of this process originates from decreasing the surface energy of small Ag3Sn particles and primary Ag3Sn [23]. Finally, Ag atoms react with the residual Sn during the cooling process, forming finely dispersed Ag3Sn particles instead of primary Ag3Sn, shown in Fig. 4d.

Fig. 4
figure 4

Illustration of bulk-type Ag3Sn IMC formation. a Back-scattered electron image of Cu/Ni/Sn3.5Ag/Ni/Cu (400 s), b The segregation of primary Ag3Sn c Formation of bulk-type Ag3Sn, d Formation of finely dispersed Ag3Sn particles

Full size image

After investigating the formation mechanism of bulk-type Ag3Sn, we conducted shear tests under four conditions to study the correlation between bulk-type Ag3Sn and mechanical reliability, Cu/Ni/Sn/Ni/Cu (200 s), Cu/Ni/Sn-3.5Ag/Ni/Cu (200 s), Cu/Ni/Sn/Ni/Cu (400 s), and Cu/Ni/Sn-3.5Ag/Ni/Cu (400 s), respectively. Fifteen samples were tested under each condition, and the results are presented in box plots as shown in Fig. 5. The average peak force of Cu/Ni/Sn3.5Ag/Ni/Cu samples increased by approximately 7% and 32% after 200 and 400 s reflow, respectively, compared to Cu/Ni/Sn/Ni/Cu samples with the same reflow time. In Cu/Ni/Sn/Ni/Cu samples, even with the increase in the IMC volume fraction after 400 s reflow, the average peak force only increased by 11% due to the formation of voids. In Cu/Ni/Sn-3.5Ag/Ni/Cu samples, bulk-type Ag3Sn were formed after 400 s reflow, and the average peak force increased by 37% as compared with Cu/Ni/Sn/Ni/Cu after 400 s reflow. It is revealed that Ag3Sn exhibits superior ductility, and as the load increases, there is a significant increase in the quantity of slip bands within the grains [24]. The superior E/H value of Ag3Sn compared to that of Ni3Sn4, Cu6Sn5, and Cu3Sn indicates that IMCs can more effectively absorb energy during the deformation process, thereby enhancing the impact resistance performance of the joints [24, 25]. As compared to the presence of voids leading to stress concentration, the formation of Ag3Sn filling the voids may effectively enhance the mechanical reliability of microbumps. In addition, the impact resistance of the joint is also enhanced. Generally, there is a positive correlation between strength and hardness. In this study, via the addition of Ag to the solder, the effect of Ag3Sn filling voids significantly increased the average peak stress by 37%. The role of Ag3Sn IMC in the mechanical reliability of Ni/Sn joints mainly lies in absorbing impact energy. Compared with the presence of void leading to stress concentration, Ag3Sn filling voids more effectively enhances the mechanical reliability of microbumps. This study suggested that adding Ag to solder can lead to greater energy absorption capacity to resist external stresses, exhibit better fracture toughness, and further improve shear strength.

Fig. 5
figure 5

The a shear strength and b end energy of Cu/Ni/Sn/Ni/Cu (200 s), Cu/Ni/Sn-3.5Ag/Ni/Cu (200 s), Cu/Ni/Sn/Ni/Cu (400 s) and Cu/Ni/Sn-3.5Ag/Ni/Cu (400 s)

Full size image

To further investigate the positions of the weak points in bumps, the fracture path was observed in Fig. 6. Since Sn possessed lower strength and was more deformative compared to IMCs, cracks tended to propagate through residual Sn in the Cu/Ni/Sn-3.5Ag/Ni/Cu (200 s) and Cu/Ni/Sn-3.5Ag/Ni/Cu (200 s) samples. We focused on the samples with samples with 400 s reflow time. With the full (Ni,Cu)3Sn4 structure, the cracks propagated along the voids which demonstrated relatively smooth fracture paths, as shown in Fig. 6a. It indicated that these voids may become weak points in the bumps. As the addition of Ag to solder, the formation of bulk-type Ag3Sn caused the microstructure undergoes a significant change. In the Cu/Ni/Sn-3.5Ag/Ni/Cu (400 s) samples, it is observed that bulk-type Ag3Sn can effectively inhibit crack propagation, as shown in Fig. 6b. Because Ag3Sn possesses excellent toughness, it allows bumps to absorb more external forces during crack propagation. Consequently, while voids act as critical points for stress concentration and facilitate crack propagation in shear test, the formation of bulk-type Ag3Sn in Cu/Ni/Sn-3.5Ag/Ni/Cu bumps occupies the original void sites, leading to more complex crack paths, thereby enhancing the overall reliability of the joints.

Fig. 6
figure 6

The cross-section images of a Cu/Ni/Sn/Ni/Cu (400 s) b Cu/Ni/Sn-3.5Ag/Ni/Cu (400 s) and schematic illustration for fracture mechanism of c Cu/Ni/Sn/Ni/Cu (400 s) d Cu/Ni/Sn-3.5Ag/Ni/Cu (400 s)

Full size image

4 Conclusion

Microstructure and mechanical performance of four sets of microbumps were investigated in this study, Cu/Ni/Sn/Ni/Cu (200 s), Cu/Ni/Sn/Ni/Cu (400 s), Cu/Ni/Sn-3.5Ag/Ni/Cu (200 s), and Cu/Ni/Sn-3.5Ag/Ni/Cu (400 s). Through the shear test results, Cu/Ni/Sn-3.5Ag/Ni/Cu (400 s) demonstrated superior strength and toughness compared to the others. The strengthening mechanism was mainly attributed to the formation of bulk-type Ag3Sn, and the fracture path views were also used to clarify the failure mode. In Cu/Ni/Sn-3.5Ag/Ni/Cu (400 s), the presence of bulk-type Ag3Sn can occupy the original site of voids, due to its excellent toughness, enabling microbumps to absorb more energy when external force is applied, thus enhancing the mechanical performance. In this study, the role of Ag3Sn IMC differs from its role on BGA scales, by adding Ag into the solder, the formation of bulk-type Ag3Sn can enhance the reliability of microbumps, making it a potential candidate for improving the mechanical reliability of 3D-IC packages.