High reliable and high bonding strength of silver sintered joints on copper surfaces by pressure sintering under air atmosphere

The continuous demand for power electronics packaging with higher power density, higher voltage and smaller size requires the replacement of silicon by wide band gap semiconductor devices such as silicon carbide and gallium nitride [1]. Accordingly, substrates with high thermal conductivity and good mechanical stability are required for the high power electronic devices. Aluminum nitride is widely used as a ceramic substrate for high power devices because its thermal conductivity is higher than conventional alumina substrate. However, the poor mechanical properties of aluminum nitride make it not suitable to use in harsh environments. In recent years, silicon nitride is considered as an alternative ceramic substrate due to its higher thermal conductivity compared to alumina and in addition its mechanical properties are better than aluminum nitride [2, 3].

To further improve the reliability of high power electronics packaging, interconnect material which is able to use at high operating temperature and with excellent properties such as high thermal and electrical properties is required. Several studies [4–6] conducted in recent years have demonstrated that silver sinter materials to be highly reliable interconnect materials for die attach application especially in power electronics devices with demanding requirements such as high efficiency and longer lifetime. Our previous studies [5, 7, 8] have demonstrated the feasibility of semiconductor devices attachment on silver and gold surfaces by either non-pressure or pressure sintering. Investment in processes and cost remains challenging. Die attachment on bare Cu surfaces has attracted increasing attention due to its potential in reducing manufacturing cost since additional metallization layer such as Ag and Au is not needed. However, presently, sintering on Cu surfaces is still a challenge. The main problem is oxidation of copper increases exponentially when exposed to air atmosphere and high temperature, which is exact the process during the sintering process. The oxide formed decreases the bonding strength between silver and copper thereby weakening the silver sinter joint. Sintering performed under inert or reducing atmosphere can obtain good bonding quality on copper surfaces as reported by recent studies [9–11]. Nevertheless, sintering performed under reducing atmosphere is challenging for a high volume production environment.

In this study, a sinter paste was developed for pressure sintering which is able to sinter on precious metal surfaces as well as on bare Cu surfaces under air atmosphere. To evaluate the sintered joint reliability, temperature cycling test and high temperature storage were performed on the sintered samples. Characterization methods consisting of die shear test, bending test, scanning acoustic microscopy and scanning electron microscopy were used to evaluate the bonding strength and delamination of the sintered joints.

ASP 338-28 sinter paste was developed for pressure sintering process on precious metal surfaces (Ag and Au) and on bare copper surfaces under air atmosphere. 0.3 mm Cu thickness Si₃N₄ active metal brazed copper (AMC) substrates with Ag metallization (different Ag metallization thickness: 0.05 μm, 0.15 μm and 0.3 μm) and without metallization (bare Cu) from Toshiba Materials were used in this study. For die attach, we used Ag metallized mechanical silicon dies with sizes of 4 mm × 4 mm and 10 mm × 10 mm. Additionally, 8 mm × 8 mm Ag metallized functional Diode from Infineon was used.

ASP 338-28 sinter paste was printed onto Si₃N₄ AMC substrates with Ag metallization and without metallization using stencils with a thickness of 75 μm. Subsequently, the printed sinter paste was dried in a convection oven at 120°C for 20 min under N₂ atmosphere with 50 ppm O₂. For die placement, the substrate was positioned on a heating plate which was heated at 130°C and the die was then placed on the dried sinter paste with a placement force of 400 g for 2 seconds. After die placement, pressure sintering process was performed in a sinter press at 230°C with a pressure of 10 MPa for 3 min under air atmosphere.

Fig. 1 shows the example of sintered samples prepared for different characterization measurements. 8 Ag metallized mechanical silicon dies with a size of 4 mm × 4 mm were attached to an AMC substrate for die shear measurement as shown in Fig. 1(a). 1 Ag metallized mechanical silicon die with a size of 10 mm × 10 mm was mounted to an AMC substrate for bending test and scanning acoustic microscopy measurements as illustrated in Fig. 1(b). For thermal resistance measurement, 1 Ag metallized functional Diode with a size of 8 mm × 8 mm was attached on an AMC substrate and after sintering a Cu/Al wire with an area fraction of approximately 70/30 (CucorAl wire from Heraeus) was bonded on the 8 mm × 8 mm Diode (Fig. 1(c)).

It is important to determine the thermal resistance, R_th for power electronic packages as it is related to thermal management of the packages. In this study, the thermal resistance measurement was performed using a thermal impedance measurement equipment. The equipment was used to measure the temperature different from junction to case for a simple package as shown in Fig. 1 (c). Fig. 2 shows the schematic diagram of cross section showing the thermal path between junction (chip) and case (heat sink). As shown in Fig. 2, the package is thermally connected to the heat sink using a thermal interface material. The thermocouple is positioned 2 mm below the package to measure the case temperature inside heat sink. The junction temperature is determined using the properties of chip.

For the thermal impedance measurement, a defined measurement current is applied for a short time to the test package. The cooling down behaviour is measured when the current is turned off. The thermal resistance of the package is calculated using the steady state of the thermal impedance curve based on the temperature difference between the max junction temperature of the chip (Tj) and the temperature of the heat sink (Tc) in relation to the electrical power applied as shown in Equation 1. In addition, Equation 1 also shows the transient thermal resistance (thermal impedance), Z_th is the time dependent thermal resistance [12].

where T_j is junction temperature, T_c is heat sink temperature, P_v is power dissipation

In this study, we measured the thermal resistance of the sintered samples (AMC package) before the long term reliability tests. The results were compared with a direct copper bonding (DCB) package (DCB substrate, ASP 338-28 sinter paste, 8 mm × 8 mm Ag metallized functional Diode and CucorAl wire).

Temperature cycling test (TCT) with a condition of −40/+150°C and high temperature storage (HTS) at 250°C were carried out on the sintered samples to evaluate the reliability of silver sintered joints on Ag metallized and bare Cu AMC substrates. Die shear measurements were performed on the sintered samples before and after TCT as well as after HTS to measure variations in bonding strength of silver sintered joints. The die shear measurement is a standard test method to determine the shear strength of bonding materials. It is based on a measure of force applied to a semiconductor die mounted to a substrate using sinter paste as a bonding material. In this study, Nordson Dage 4000 plus was used to measure the bonding strength of silver sintered joints. Fig. 3 shows the schematic diagram of die shear measurements and different die shear failure modes. Three common die shear failure modes are shown in Fig. 3: (1) adhesive break on the die, (2) adhesive break on the substrate and (3) cohesive break in the sinter layer. Cohesive break in the sinter layer is usually observed for sintered joints with high bonding strength. In general, high die shear strength > 30 N/mm² is achieved for this type of failure mode. In contrast, adhesive break either on the die or on the substrate is observed for sintered joints with low bonding strength and the die shear strength is low.

Scanning electron microscopy (SEM) was used to characterize the microstructure of silver sintered joint. The AMC substrate with 0.05 μm Ag metallization thickness sintered samples before and after 2000 thermal cycles were characterized by SEM.

Fig. 4 shows the schematic diagram of bending test. Bending test was performed on the sintered samples before and after 2000 thermal cycles. Bending test is a rather quick method to examine the bonding strength of sintered joints. As shown in Fig. 4, for low bonding strength, die peeled off after bending test, while, die still attached on substrate for high bonding strength.

Scanning acoustic microscopy (SAM) is a non-destructive method. SAM was used to measure the sintered samples before and after TCT to identify voids, drying channels and delamination in the silver sintered layers. Voids free is one of the important criteria for good sinter joints because voids created in the sinter layer would hinder the heat dissipation and as a result become a hotspot [13].

Thermal resistance of the initial sintered samples was measured by thermal impedance measurement equipment as described in section II. C. It is important to note that R_th changes with the nature of thermal interface material (TIM). Therefore, we can only compare R_th of packages when identical equipment and TIM are used to perform the measurement. Fig. 5 illustrates the R_th for the sintered samples before the long term reliability tests. It can be seen from Fig. 5 that the Rth is rather similar for all the samples which is approximately 0.8 K/W. In addition to AMC packages, R_th of DCB package was also measured. R_th of about 0.95 K/W was obtained for DCB package which is slightly higher than for AMC package. As mentioned in section I, the thermal conductivity of Si₃N₄ is higher than Al₂O₃ and this explained the lower thermal resistance of AMC package.

Sintered samples with 4 mm × 4 mm Ag metallized dies as shown in Fig. 1(a) were sheared to examine the bonding strength of sintered joints. Die shear measurement was performed on the sintered samples before and after TCT and the die shear strengths are illustrated in Fig. 6. 24 dies were sheared to generate an individual boxplot.

It can be seen from Fig. 6 that the average initial die shear strength of 33 N/mm² for Ag metallized AMC substrates is higher than for bare Cu AMC substrate (13 N/mm²). This phenomenon can be explained by the diffusion between Ag and Ag is much faster than the interdiffusion between Ag and Cu. Similar observation also reported by Bai [6]. Furthermore, an interesting phenomenon was observed for all the sintered samples in which the average die shear strength increased significantly after 1000 and 2000 cycles. It is believed that the sintering process is not yet completed under the sintering process conditions used in this study. As a result, diffusion of the Ag and Cu occurred during thermal cycling process and consequently strengthen the sintered joints. The average die shear strength for AMC with different Ag metallization thickness (0.05 μm, 0.15 μm and 0.3 μm) are rather similar demonstrating that the thickness of Ag metallization has no significant effect on the bonding strength of sintered joints. It is important to point out that after 2000 cycles the average die shear strength for bare Cu substrate increased to 44 N/mm² which is relatively similar to the average die shear strength for Ag metallized substrates.

Besides TCT, HTS was performed on the sintered samples (Ag metallized AMC with Ag thickness of 0.05 μm and bare Cu AMC). Fig. 7 shows the die shear strength before and after HTS. The results also show that the average die shear strength increased after HTS. Similar explanation also applied to this phenomenon which is Ag and Cu continued to diffuse during HTS and resulting in higher bonding strength. As can be seen in Fig. 7 that there is no significant difference in the average die shear strength between 250 h, 500 h and 1000 h storage. It is strongly believed that after a certain time of storage at 250°C diffusion of Ag and Cu discontinue and sintering process is completed. As a result, there is no further increase in bonding strength and the die shear strength remained constant.

As described in section II. C, three common die shear failure modes are observed for silver sinter joints. Fig. 8 shows the die shear failure mode before and after 2000 thermal cycles. It can be seen from Fig. 8 that adhesive break on the substrate was observed for bare Cu substrate before TCT where silver sintered layer was only formed on the die back side with Ag metallization. In contrast, a mixture of adhesive and cohesive break was observed for Ag metallized substrates where silver sintered layer was found on both the die back side and the Ag substrate surface. This phenomenon demonstrates that Ag diffuses faster into Ag than into Cu. As mentioned in section II. C, die shear failure mode and die shear strength are strongly related to each other. Initial sample of bare Cu substrate achieved adhesive break failure mode explained its low die shear strength (13 N/mm²). After 2000 thermal cycles, cohesive break in the sintered layer was observed for all the samples and this observation further verify the increased die shear strength.

Fig. 9 shows the die shear failure mode before and after HTS. It is worth noting that cohesive break in Cu layer was observed for the samples stored for 1000 h. It is likely that Cu from the substrate diffused into silver sintered layer and simultaneously Ag from the sintered layer diffused into the Cu layer of substrate during HTS. It has been reported that the diffusion rate of Cu to Ag is higher than that of Ag to Cu [14].

Watanabe et al. [15] conducted pressureless silver sintering on bare Cu substrate and via elemental analysis they observed an interdiffusion between Ag and Cu occurred. Their analyses are in agreement with literature illustrating the diffusion rate of Cu to Ag is higher than that of Ag to Cu and as a result, it is strongly believe that the sintered joint between Cu and Ag was driven by the diffusion of Cu to Ag rather than Ag to Cu. The cohesive break failure mode obtained for the samples after 1000 h storage explained the increased die shear strength.

Fig. 10 shows the SEM cross-sectional images of AMC substrate with 0.05 μm Ag metallization thickness sintered samples before and after TCT. It can be seen from the SEM images that the silver sintered layer after 2000 thermal cycles (Fig. 10 b) is denser and more homogeneous than the silver sintered layer before TCT (Fig. 10 a). As mentioned previously, it is strongly believed that Ag continued to diffuse during TCT and it diffuses from high concentration area to low concentration area. As a result, the sintered layer became more homogeneous and additionally increased the density of silver sintered layer. This observation explains the increase of bonding strength of silver sintered joint after TCT. Furthermore, Fig. 10 shows that delamination of silver sintered joint was not observed after 2000 thermal cycles indicating that high reliable silver sintered joint was obtained by pressure sintering under air atmosphere.

Bending test is an alternative method to evaluate the bonding strength of sinter joint. Bending test was performed on the samples (substrate with 0.15 μm Ag metallization thickness and bare Cu substrate) before and after 2000 thermal cycles and the results are shown in Fig. 1. For both samples, similar results are obtained before and after TCT in which the die is strongly attached on the substrate after bending test demonstrating that high bonding strength of sintered joint with no delamination in the sintered layer was achieved. The bending test results further confirm that high bonding strength of sintered joint was obtained after 2000 thermal cycles.

SAM measurement was performed on the samples before and after 2000 thermal cycles and the images are shown in Fig. 2. SAM images show that voids and drying channels were not detected in the sintered layer for the samples before TCT and additionally homogeneous sintered layer was obtained indicating that homogeneous pressure distribution was applied during sintering process under the sintering conditions used in this study. Furthermore, voids and drying channels were not observed for the samples after TCT. It is important to point out that delamination in the sintered layer for the samples after 2000 thermal cycles was not identified by SAM.

This study demonstrated that with a properly formulated sinter paste it is possible to produce high reliable and high bonding strength of silver sintered joints on copper surfaces by pressure sintering under air atmosphere. Our thermal resistance measurement indicates that Si3N4 substrate has lower thermal resistance than Al₂O₃ substrate. We observed that the average initial die shear strength for Ag metallized substrates is higher than for bare Cu substrate. This phenomenon is attributed to the diffusion between Ag and Ag is faster than between Ag and Cu. This explanation is verified by the die shear failure mode. A mixture of adhesive and cohesive break was observed for Ag metallized substrates before TCT and HTS. In contrast, adhesive break on the substrate was observed for Cu substrate. The die shear results also demonstrated that the thickness of Ag metallization has no significant effect on the bonding strength of sintered joints. The average die shear strength increased significantly after TCT and HTS. It is believed that Ag and Cu continued to diffuse during TCT and HTS and consequently strengthen the sintered joints resulting in higher bonding strength. SEM cross-sectional images shows that the silver sintered layer after 2000 thermal cycles is denser and more homogeneous than that before TCT. Cohesive break was observed for the samples after TCT and HTS and this observation explained the high die shear strength. The average die shear strength for Cu substrate after 2000 thermal cycles is relatively similar to that obtained for Ag metallized substrates. No further increase in the die shear strength was observed after 250 h storage at 250°C illustrating that after a certain time of storage the diffusion of Cu and Ag discontinue and the sintering process is completed. The bending test results further confirm that high bonding strength of sintered joints was achieved for samples after 2000 thermal cycles. SAM images show that voids, drying channels and delamination in the sintered layer were not observed for the samples after 2000 thermal cycles. Furthermore, SEM cross-sectional image illustrates that delamination of sintered joint was not observed after 2000 thermal cycles.