Reliability and prediction of Sn36Pb2Ag solder joints under thermal aging test

In this work, the capacitor solder joints were aged at 50 °C, 75 °C, 100 °C, 125, and 150 °C from 100 h to 1000 h. The intermetallic compounds (IMCs) layer growth of Sn36Pb2Ag on hot air solder leveled (HASL) pad and electroless nickel/immersion gold (ENIG) pad was measured. Based on the empirical power-law of the IMC growth and the Arrhenius relationship between diffusion coefficient and aging temperature, a method to predict the IMC growth at a selected temperature was developed. The mechanical property of capacitor solder joints after thermal aging was investigated through the shear test. Through analysis of the fracture surface, the mixed fracture mode of ductile and brittle was exhibited. The porous structure of the Cu coating on the capacitor electrode was determined to be the origin of the crack.


Introduction
With the development of electronic devices towards the direction of miniaturization and multifunctionalization, many advanced packaging technologies, such as chip size package (CSP) and system in package (SiP) have developed in the field of electronic packaging [1]. In recent years, the emergence of through Si via (TSV) and wafer level package (WLP) has driven the development of 3D microelectronic packaging [2]. Solder joints provide mechanical support and electronic connection for electronic components, which will meet greater thermal and mechanical challenges in 3D packages [3]. Due to the harm of Pb to the environment and human health, legislation has banned Pb usage in commercial electronic products [3]. A series of lead-free solders such as SnAg, SnAgCu, SnCu, SnZn, and SnBi solders have been invented to replace the traditional Pbcontained solders [4]. Although these lead-free solders have superior performance in some aspects, many mechanical potential risks, such as high melting point and low wettability, have been exposed [5]. The solders with high melting points need a higher soldering temperature which will introduce huge thermal stress during packaging. In contrast, the Pb-contained solders have excellent mechanical properties. For reliability consideration, Pb-contained solders are still used in highly precious devices [6]. In particular, long-term reliability is assumed as the most important issue for solder joints in military and aerospace industries. Thus, the Sn36Pb2Ag solder was applied in this study.
IMC layer formed at the interface of the metal substrate and solder during soldering plays an essential role in the mechanical strength of solder joints. An appropriate thickness of the IMC layer is necessary to guarantee the soldering quality. The effects of the process temperature profile, solder, and metal finish on the initial thickness of solder joints have been extensively researched [7][8][9]. Due to solid diffusion at the interface of solder and substrate, the growth of IMC layer still exists after the reflow process. The diffusion of the atom will be accelerated by the high temperature. In addition, excessive growth of the IMC layer during thermal aging will degrade the strength of solder joints owing to the intrinsic brittleness of IMC. Tang et al [10] reported the mechanical properties degradation of the solder joint on the effect of annealing temperature. The thermal aging test was also used to investigated the constitutive, creep, and fatigue behavior of the solder alloy [11][12][13][14]. A Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
preservation layer, such as Ni layer coating on the substrate is regularly used to prevent the interfacial reaction of solder and substrate in electronic packaging. Element doping in the solder or Cu substrate has also been proposed to restrain the growth of IMC. Scholars recently have shown the effective inhibition of IMC growth with the addition of Ni, Zn, and Co in lead-free solder or Cu substrate [15][16][17][18]. Furthermore, the growth kinetics of IMC as the more fundamental issue has been assumed to match the growth kinetics of IMC. During isothermal aging, the growth kinetics of IMC is dominated by volume diffusion and follows the assumption that the IMC thickness is in direct proportion to the square root of aging time [19]. Meanwhile, the IMC thickness is deeply influenced by aging temperature which can be controlled by the Arrhenius relationship [20]. Ke et al [21] investigated the effects of temperature on the diffusion coefficient of Cu atom in Sn.
In electronic packaging, capacitors as the most commonly used passive component, are directly mounted on the substrate to maintain the voltage across all operating frequencies [22]. The reduction of capacitor dimension results in the challenge of handling and assembling capacitors [23]. In this case, the reliability of capacitor solder joints is more worthy of study. Passing test is mostly used for traditional reliability evaluation. However, this cannot predict the service life of the electronic products. In this work, we have conducted the thermal aging test at a different temperature to observe the IMC growth kinetics in capacitor solder joints with two types of surface finish. The temperature effects on the IMC growth kinetics have been fitted with the Arrhenius relationship aiming to give a prediction of IMC thickness at a selected temperature. Especially at low service temperature or storage temperature, in which the IMC grows slowly, the prediction of IMC growth can greatly save the cost of time. The IMC layer thickness was a crucial factor for the reliability of the solder joints. Therefore, we can estimate the service life based on the growth of IMC layer thickness. Then, a shear test was implemented to assess the mechanical property of Sn36Pb2Ag solder joints after thermal aging.

Prediction model
The growth kinetics of IMC thickness under thermal aging is summarized as the power-low relationship [24], In which h t and h 0 are the average thickness of IMC layer at time t and initial time, respectively. k and n donate the diffusion coefficient and time exponent. This empirical model has been combined with the actual IMC growth mechanism [25][26][27]. In the present study, the growth of IMC is mainly controlled by the volume diffusion where the time exponent n = 0.5. In this work, we have obtained the IMC growth dates at five different temperatures. Hence, five sets of diffusion coefficient and temperature can be calculated by the formula (1).
The Arrhenius model describes the relationship between diffusion coefficient k and temperature during the diffusion process [20].
Where k 0 is the pre-exponential coefficient, Q is the activation energy, R is gas constant (8.314 J mol −1 K −1 ), and T is the absolute temperature. Rewrite formula (2) as follows, Based on the data sets of the diffusion coefficient and temperature, we can calculate the uncertain parameters (Q and k 0 ) in the Arrhenius model by the formula (3). According to the obtained Arrhenius model, we can calculate growth rate constant k at a selected temperature. It should be noted that the IMC growth still follows the parabolic law at the selected temperature. If the initial IMC thickness is obtained, the thickness of IMC at selected temperatures with different aging times will be predicted by formula (1).

Experimental
The tested printed circuit board (PCB) adopts two kinds of surface finish, namely HASL and ENIG. HASL pad is immersed with molten SnPb alloy on the surface of the Cu pad. As shown in figure 1, the initial thickness of SnPb layer on HASL pad is uneven, and the thickest part of SnPb layer can reach 135 μm. ENIG pads contain two coating layers. The surface is a 0.15 μm thick Au plating layer, Ni and P are usually combined to form amorphous alloys seen in figure 2. The pad and capacitor were soldered together using Sn36Pb2Ag solder. The samples were put in the drying box (WGLL-65BE) for 100 h to 1000 h solid-state aging. Test temperature was set to 50°C, 75°C, 100°C, 125°C, and 150°C. Then the samples were taken out and encapsulated with epoxy resin. The cross-section of the sample was ground until no obvious deep scratches under the optical microscope with 240 #, 800 #, 1500 #, 2000 #, and 3000 # silicon carbide sandpaper respectively, and then polished using 0.5 μm diamond polishing agent. The solder joint samples were observed by scanning electron microscope (SEM, supra 55vp, Zeiss, Germany) equipped with an energy dispersive spectrometer (EDS). The composition of IMC at the solder joint interface is determined by EDS, and the IMC organization photos are taken in backscattering mode. The SEM image of interface was digitally processed by Adobe's Photoshop software to measure the thickness of interface IMC [28]. The thickness of IMC layer is calculated by the following formula, Where X is the average thickness of IMC layer, A and L represent the number of pixels in the area and length of IMC layer, respectively. P is the actual length of a unit pixel. The thickness for each experimental condition was taken from the average of more than three measurements. Each measurement was a random selection at the interface of the same solder joint. Destructive shear tests were performed on capacitor solder joints using a static shear force tester. The fixture and sample are placed in a temperature control box to maintain the test temperature. The shear speed and shear height were set to 200 μm s −1 and 500 μm, respectively.

Results and discussion
Analysis of interface IMC composition The interface of HASL pad and capacitor solder joint is regarded as the interface of Sn36Pb2Ag/Cu. Figure 3 and   reported in the literature [29]. For the ENIG pad, it is considered as the interface of Sn36Pb2Ag/Ni. A thick Au layer (the thickness of Au layer is generally greater than 2 ∼3 μm) will form the AuSn 4 IMC in the solder matrix during thermal aging, which causes the brittleness of solder joint [30]. In this study, the Au layer (0.15 μm) is so thin that can be ignored in the following study. From figure 4 and table 1, the interface IMCs are identified as Ni 3 Sn 4 and Ag 3 Sn. With the consumption of Ni during aging, a thin layer of Ni 3 P has existed between the Ni(P) and Ni 3 Sn 4 , which has been reported in previous research [4].

Growth of interface IMC
The average thickness of IMC layer for HASL and ENIG pads are listed in tables 2 and 3 respectively. The relationship between IMC thickness with time and temperature is plotted in figure 5 and 6. It is noteworthy that at the same aging temperature (50°C and 75°C), the IMC thickness with a short aging time is thicker than that with a long aging time. For example, the IMC thickness of HASL-solder joint under 50°C for 250 h is higher than the 500 h and 1000 h. It can be attributed to the difference between the initial IMC layer thickness of all solder joints. Especially at low aging temperatures, the growth rate of IMC is low, and the total thickness      increment of IMC layer was not more than 0.5 μm. Therefore, the solder joint with a thicker initial IMC layer may be a singularity of the experimental data. In the process of solid-state aging, the growth rate of the interface IMC layer thickness decreases with the increase in thickness [31]. At low aging temperatures, the growth rate of IMC layer was greatly affected by the initial thickness of IMC layer. Hence, the singularity of the date was abandoned for the following calculation. While at a high temperature, the IMC growth was high enough to ignore the effect of the initial difference. Compared with these two kinds of solder joints, the IMC growth rate of HASL solder joint is faster and shows a strong correlation with the temperature. While the IMC of ENIG solder joint depicts a relatively slow growth rate even at high aging temperatures. In order to fit the parabolic law accurately, the thickness of IMC layer is properly corrected based on the above analysis. The fitting curve of the IMC thickness versus the square root of aging time at different temperatures is displayed in figures 7 and 8. The slope k of the curve increases with the rise of temperature. The curve fitting degree at high temperature is better than that at low temperature. This is because the IMC growth at low aging temperatures is greatly affected by the initial IMC thickness. The fitting curve of diffusion coefficient k with temperature T is shown in figure 9. The activation energy is calculated as 90293.63 J mol −1 and 64425.87 J mol −1 on HASL pad and ENIG pad, respectively. The difference in activation energy can be attributed to the variety of IMC layers [32]. For aerospace and military fields, electronic components need high reliability and long service life. Electronic products need to be in service for several years or even decades at room temperature or lower service temperature. The prediction of IMC growth of solder joints with low temperatures and long service times is crucial to evaluate the reliability of electronic components. Based on the above methods, we calculated that the growth thickness of IMC at room temperature for 10 years is 0.202 μm and 0.402 μm for the HASL and ENIG solder joints, respectively. The IMC layer of ENIG solder joints grows faster than the IMC layer of HASL solder joints. According to the Arrhenius model, the diffusion coefficient for ENIG solder joints (5.13e −22 m 2 s −1 ) is larger than the HASL solder joints (1.16e −22 m 2 s −1 ) at room temperature. Therefore, a thicker IMC layer was predicted of the ENIG solder joint at room temperature for 10 years. This method is also applicable to the prediction of IMC growth of other solder joints. It is worth noting that it is necessary to collect IMC growth data under the real service state of the product to verify and modify the method. This is exactly what we need to do next.  Mechanical test of solder joint Figure 10 shows the shear properties at different aging temperatures and times of HASL-solder joints. After high temperature aging (150°C and 125°C), the static shear force decreases with the aging time prolonging. While at low aging temperature (below 125°C), the static shear force fluctuates at the early stage, and finally the static shear force after aging 1000 h reaches a higher level than the initial stage. Interestingly, all the fracture occurs at the capacitor side. The capacitor at the initial stage, aging at 150°C for 1000 h and aging at 100°C for 100 h (the lowest shear force) was selected to analyze the fracture surface using SEM. The fixed fracture model was observed for all fracture surfaces. From the EDS results of figure 11 and table 1, the fracture surface of the left side is identified as the surface of solder matrix and IMC layer for the existence of Sn, Pb, and Cu. In the center fracture surface, plane surface of Cu and Ni is determined as the capacitor electrode, and the surface with dimples is the  solder matrix. Ductile and brittle fractures are assumed to happen at the left side solder matrix and the capacitor electrode surface, respectively. The cracks originate from the capacitor electrode and propagate to the solder matrix. After aging at 150°C for 1000 h, the fracture surface is similar to the initial stage as shown in figure 12. The variation is the surface of solder matrix performs little dimples and more cleavages. This can be attributed to the coarsening of Pb phase under thermal aging which improves the brittleness of solder matrix. Figure 13 describes the total brittle fracture surface through the capacitor electrode of the solder joint aging at 100°C for 100 h. Based on the above analysis, the surface of capacitor electrode is believed as the initiation point of cracks   under shear test. The EDS results are introduced in figure 14 and table 1 for the surface finish of capacitor electrode. From the right to lift, it is Cu and Ni coating layer, Ni 3 Sn 4 IMC layer, and SnPb coating respectively. It is obvious to observe a porous structure on the Cu coating layer. The porous structure is susceptible to forming the cracks under shear test, which is the reason for the brittle fracture at capacitor electrode side.

Conclusion
In this paper, the thermal aging test and shear test are carried out on the capacitor solder joints with the HASL and ENIG surface finish. The main conclusions are listed below: 1. The IMC thickness was fitted with the empirical power-law model under different aging temperatures. The activation energy was calculated as 90 kJ mol −1 and 64 kJ mol −1 on HASL pad and ENIG pad, respectively.
2. Based on the Arrhenius relationship, we calculated that the growth thickness of IMC at room temperature for 10 years is 0.202 μm and 0.402 μm for the HASL and ENIG solder joints, respectively. The IMC layer of ENIG solder joints grows faster than the IMC layer of HASL solder joints which is due to the larger diffusion coefficient of ENIG solder joints at room temperature.
3. The mixed fracture mode of ductile and brittle was founded in the solder joints during the shear test. The porous structure of the Cu coating on the capacitor electrode was determined to be the origin of the crack which is the failure point for the solder joints.