Effect of Ar Plasma Treatment for Surface of Insert Metal on Property of TLP Bonding Joint for Power Module

Seungju Baek; Dong-Yurl Yu; Jun-Hyuk Son; Junghwan Bang; Jungsoo Kim; Min-Su Kim; Han-Bo-Ram Lee; Yong-Ho Ko

doi:10.5781/JWJ.2020.38.4.5

Abstract

In this study, we investigated the effect of plasma treatment for the surface of an insert bonding metal on the properties of a transient liquid phase (TLP) bonding joint for a power module. Sn preforms, as the insert bonding metal for the TLP bonding, were used with and without Ar plasma treatment. To investigate the effect of Ar plasma treatment, the TLP bonding for two structures of Cu-finished Si chip/Sn preform/organic solderability preservative (OSP)-finished direct bond copper (DBC) substrate (Cu/Sn/Cu), and Ni-finished Si chip/Sn preform/Ni(P)-finished DBC substrate (Ni/Sn/Ni) was performed with 1 MPa at 300 °C and the bonding times were 10, 30, and 60 min, respectively. After the TLP bonding, we observed interfacial reactions and formations of intermetallic compounds (IMCs) under various bonding conditions. To evaluate mechanical properties, a shear test was also performed. Compared to the TLP bonding joint that used bare Sn preforms without Ar plasma treatment, growth of IMCs at the bonding joint that used Sn preforms with Ar plasma treatment occurred faster, and the IMCs could be formed through the entire joint despite the bonding time of 10 min. Meanwhile, by increasing the bonding time, Cu₆Sn₅ and Cu₃Sn were formed at the Cu/Sn/Cu TLP bonding joint, whereas Ni₃Sn₄, Ni-Sn-P, and Ni₃P were observed at the Ni/Sn/Ni joint. In the case of the Cu/Sn/Cu joint, we observed that increasing Cu₃Sn formation while increasing the bonding time could be beneficial to the shear strength of the joint. Further, shear strengths of the joint were not significantly changed under the bonding conditions after Ni₃Sn₄, Ni-Sn-P, and Ni₃P IMCs were formed at the entire joint of the Ni/Sn/Ni.

Key words: Transient liquid phase (TLP) bonding, Power module, Interfacial reaction, Intermetallic compound (IMC), Shear test

1. Introduction

The power module, which serves to convert, transform, distribute, and control electrical energy, is a core component of inverters and converters in fields such as renewable energy, automobiles, and railways¹⁾. The power module consists of a power semiconductor chip, substrate, Al or Cu wire for electrical connection, base plate, etc., and functions as a module through the bonding of each component. The use of power modules has been gradually expanding recently, and eco-friendly vehicles such as pure electrical vehicles (EVs) and hybrid electrical vehicles (HEVs) are being increasingly used²⁾. Si, SiC, and GaN-based devices are used for the power semiconductor chip constituting the power module³⁾, and ceramic-based substrates such as direct bond copper (DBC) and active metal brazing (AMB) are used for the substrate through various surface treatments⁴⁾.

For the bond between the power semiconductor chip and substrate, a soldering technology using solders containing Pb has been used for some time. However, as recent environmental regulations have prohibited the use of Pb, Sn-based lead-free solders (e.g., Sn-Ag-Cu, Sn-Zn) have primarily been used⁵⁾. As power modules used in automobiles are generally exposed to high-temperature environments of 200°C or more, when applying the soldering method using lead-free solders based on Sn with a melting point of approximately 231 °C, reliability-related issues such as a re-melting phenomenon can occur during using the power module after bonding the chip and substrate⁶⁾. To address this problem, many bonding methods are studied for applying to power module using bonding materials and bonding processes applicable in high-temperature environments. In this regard, a soldering method using an Au-Sn lead-free solder with a high melting point⁷^,⁸⁾ and a sintering method using pastes of Ag or Cu metal powders are being studied⁹^,¹⁰⁾. However, the Au-Sn lead-free solder is difficult to industrially apply due to the high price of Au¹¹⁾. Regarding the sintering method using a metal powder such as Ag as a bonding material, though it has the advantage of improved characteristics with finer particles, material costs also increase with finer particles¹²⁾. Furthermore, when using a Cu metal powder, it is difficult to solve the oxidation problem of the Cu material itself¹⁰^,¹²⁾. Accordingly, a transient liquid phase (TLP) bonding technology has been recently proposed to improve on the disadvantages of the soldering and sintering bonding methods¹³⁾.

TLP bonding forms intermetallic compounds (IMCs) at the bonding joint, which form through the diffusion of metal atoms at relatively low temperatures, while melting the low-melting-point insert metal that acts as a bonding material between the high-melting-point material¹³⁾. The IMCs formed throughout the entire bonding joint can secure reliability even in high-temperature environments due to their high melting point. However, researchers have noted that one major disadvantage of TLP bonding is that long process times are required to form the entire bonding joint through uniform IMCs¹¹⁾_.

This study reported on the IMC formation and growth of a TLP bonding joint according to the TLP bonding time through plasma pretreatment on a metal bonding material inserted for bonding. A metal foil Sn preform was used as the bonding material and the influences of Ar plasma treatment on the interfacial reaction of the TLP bonding joint, IMC formation, and mechanical properties through Ar plasma treatment on the surface of Sn preform were investigated.

2. Experimental Method

Si chips and DBC substrates were used for TLP bonding. For the Si chip, Ti and Cu were deposited to 0.05 and 0.5 μm on the Si wafer via sputtering, respectively. To then compare the properties of the TLP bonding joint according to different final surface-finished materials of the chip, Cu and Ni were separately deposited to fabricate two types of Si wafers. The Cu and Ni were finally deposited to a thickness of 10 μm using the electroplating method. The Si wafer on which the Cu and Ni were finally deposited was diced into a size of 4 mm × 4 mm and used as the chip for TLP bonding. For the substrate used for bonding, DBC substrates in which 300 μm of Cu was stacked on the upper and lower surfaces of a 380 μm Al₂O₃-based ceramic substrate were used. This DBC substrate was 17.8 mm × 12.7 mm × 0.98 mm in size. To compare the bonding joint properties according to the surface-finished materials, two types of DBC substrates were also fabricated. One was the DBC substrate finished with organic solderability preservative (OSP) of several nm and the other was the DBC substrate treated with Ni(P) of approximately 5 μm. Meanwhile, for the TLP bonding material, we a Sn preform with a size of 4 mm × 4 mm and thickness of approximately 40 μm. To investigate the influence of Ar plasma treatment on the TLP bonding joint, Ar plasma pretreatment on the upper and lower surfaces of the Sn preform was performed. The Ar plasma surface treatment was conducted for approximately 10 minutes at a gas flow of 100 sccm, a power of 100 W, and a vacuum degree of 0.5 torr. To perform TLP bonding, the Sn preform was first placed on the DBC substrate and the chip was placed on the Sn preform. After placing, the TLP boding was performed using a chamber capable of controlling the temperature and pressure in a vacuum state. Two samples with TLP bonding were fabricated: a Cu surface-treated Si chip/Sn preform/OSP surface-finished DBC (Cu/Sn/Cu) structure, and a Ni surface-treated Si Chip/Sn preform/Ni(P) surface-finished DBC (Ni/Sn/Ni) structure. After the TLP bonding, properties of the bonding joint were compared and evaluated according to the TLP bonding joint structure. TLP bonding was performed by applying a pressure of 1 MPa in a vacuum chamber at 300°C, and the properties of the TLP bonding joints according to bonding time were compared for three bonding time conditions (10, 30, and 60 minutes). Fig. 1 shows a schematic diagram of the Cu/Sn/Cu and Ni/Sn/Ni TLP bonding process.

Fig. 1

Schematic diagrams of TLP bonding process for (a) Cu/Sn/Cu and (b) Ni/Sn/Ni

Meanwhile, after TLP bonding, to analyze the structure of the bonding joint, the bonding time, and the interfacial reaction according to the Ar plasma treatment and the cross-section of the specimen was polished and then scanning electron microscopy (SEM, Inspect F, FEI, USA) and energy-dispersive X-ray spectroscopy (EDS) were used to observe and analyze the bonding joint. Moreover, to evaluate the mechanical properties of the TLP bonding joint, a shear tester (DAGE-4000, Nordson DAGE, UK) was also used to compare and evaluate the bonding strength according to each bonding condition. The bonding strength evaluation was performed at a shear height of 25 μm and a shear speed of 200 μm/s. Further, after evaluating the shear strength, cross-sectional analyses were performed on the fractured loci using SEM and EDS for the bonding joints structure, the bonding time, and the effects of Ar plasma surface treatment were compared.

3. Experimental Results and Discussion

Fig. 2 shows the SEM micrographs of the cross-sectional analysis after Cu/Sn/Cu TLP bonding. According to the cross-sectional analysis at 10 minutes of bonding time, the IMCs formed in the bonding joint consisted of Cu₆Sn₅ and Cu₃Sn IMCs, which are known¹³⁾ to typically form in the Cu-Sn system. In addition, when using the Sn preform not applied to Ar plasma surface treatment, residual Sn was remained after forming the IMCs at the bonding joint. Conversely, in the bonding joint using the Ar plasma surface-treated Sn preform, IMCs formed throughout the entire bonding joint from 10 minutes of the bonding time. When the bonding time was increased to 30 or 60 minutes, IMCs formed throughout the entire bonding joint regardless of Ar plasma surface treatment. Fig. 3 shows the ratio of IMC types forming the Cu/Sn/Cu TLP bonding joint according to the bonding time and Ar plasma treatment. Regardless of Ar plasma treatment, as the bonding time increased, the ratio of Cu₆Sn₅ decreased and the ratio of Cu₃Sn increased. Regarding the IMCs formed in the Cu-Sn system, Cu₆Sn₅ first forms by the diffusion reaction of Cu and Sn. Afterwards, Cu₃Sn is known to form between Cu₆Sn₅ and Cu due to the diffusion of additional Cu¹⁴⁾. Therefore, Cu₆Sn₅ grew at the initial Cu/Sn/Cu bonding joint and Cu₃Sn grew at the Cu₆Sn₅/Cu interface as the bonding time increased. Meanwhile, when using the Ar plasma surface-treated Sn preform for bonding at bonding times of 10, 30, and 60 minutes (Ar(Cu)-10, 30, 60), the ratio of Cu₃Sn at the bonding joint at all bonding times exceeded that in the bonding joint formed without plasma treatment at 10, 30, and 60 minutes of bonding time (Cu-10, 30, 60). When using Ar of a representative inert gas as the plasma reaction gas, Ar plasma surface treatment was performed on the Sn preform to utilize the advantages of physical organic material removal and surface activation¹⁵⁾ rather than chemical reactions with the sample surface. In this regard, it is assumed that Ar plasma treatment could enhance the wettability, reactivity, and bonding characteristics at the interface, which result in forming uniform IMCs at the TLP bonding joint and leading to a high growth rate of Cu₃Sn due to continuously accelerate Cu diffusion. Fig. 4 shows the bonding strength according to plasma surface treatment and bonding time of the Cu/Sn/Cu TLP bonding joint. When plasma surface treatment was not applied, the bonding strength was approximately 48 MPa at a bonding time of 10 minutes (Cu-10) and a relatively higher value than the other conditions. In the bonding joint without plasma treatment at a bonding time of 30 minutes (Cu-30), the bonding strength was 24 MPa, and in that with Ar plasma treatment (Ar(Cu)-30), the bonding strength was 29 MPa. When the bonding time was increased to 60 minutes, regardless of plasma treatment (Cu-60, Ar(Cu)-60), the bonding strength tended to be increased with bonding strengths of 25 and 34 MPa, respectively. After 30 minutes of bonding time, a relatively high strength was measured when plasma treatment was applied. Fig. 5 shows the SEM micrographs of the cross-sectioned fracture loci after evaluating the bonding strength of the Cu/Sn/Cu TLP bonding joint, as well as a fracture diagram based on the SEM results. The results demonstrated that, regardless of the bonding time or surface treatment, fracture occurred on the substrate-side, which was the bottom of the bonding joint. As shown in Fig. 5 (a) and (b) Case 1, fracture of the bonding time of 10 minutes without plasma surface treatment occurred at the interface of the residual Sn and IMC and the interface between the Cu₆Sn₅ and Cu₃Sn IMCs at the bonding joint. Above mentioned, the shear strength for the 10 minutes of bonding time without surface treatment was relatively high strength. From the cross-sectional analysis of the fracture loci, it was assumed that this was attributed to the ductility characteristics which were shown at Sn-based solder bonding joints due to the influence of remained Sn after the formation of IMCs at the initial bonding joint. Thus, at the initial bonding process time, when residual Sn remained in the bonding joint, the bonding strength showed a similar trend to previous research showing high bonding strength values¹³⁾. Meanwhile, in the case of 10 minutes of bonding time with plasma surface treatment, as shown in Fig. 5 (a) and (b) Case 2, fracture occurred at the interface of Cu₆Sn₅ and Cu₃Sn. This differed from the case in which Ar plasma treatment was not applied because the IMCs formed throughout the entire bonding joint from a bonding time of 10 minutes when Ar plasma surface treatment was applied. Therefore, the bonding strength for 10 minutes of bonding time with Ar plasma surface treatment was lower than that without Ar plasma surface treatment due to the fracture at the IMC interfaces because of IMCs which were formed at the whole bonding joint. As shown in Fig. 5 (a) and Fig. 5 (b) Case 3 and 4, bonding joint fractures occurred in the Cu₆Sn₅ IMC layer and the interface between Cu₆Sn5 and Cu₃Sn when the bonding times were increased to 30 and 60 minutes, regardless of plasma treatment. As mentioned in Fig. 3, the ratio of Cu₃Sn IMCs at the bonding joint increased as the bonding time increased. Although the fracture loci were similar to each other after 30 minutes of bonding time, the increase in strength as the bonding time increased was influenced by the increase in the ratio of Cu₃Sn IMCs at the bonding joint. Cu₃Sn generally has a higher mechanical properties such as Young’s modulus etc. than Cu₆Sn₅¹⁶⁾. In this study, when the bonding time increased and Ar plasma surface treatment was applied, the ratio of Cu₃Sn increased, which result in influencing on the increase in strength of the TLP bonding joint.

Fig. 2

Cross-sectional SEM micrographs of Cu/Sn/Cu joints with various bonding times

Fig. 3

Ratio of IMC formed at Cu/Sn/Cu joints with (Ar(Cu)) and without(Cu) plasma surface treatment

Fig. 4

Shear strength of Cu/Sn/Cu joints for various bonding times and Ar plasma treatment

Fig. 5

(a) Cross sectional SEM micrographs of fracture surface and (b) fracture mode of Cu/Sn/Cu joints after shear test

Fig. 6 shows the SEM micrographs of the cross-sectioned joints after Ni/Sn/Ni TLP bonding. The IMCs formed at the bonding joint were Ni₃Sn₄. When Ar plasma surface treatment was not applied, residual Sn was observed in the middle of the bonding joint for the bonding time of 10 minutes as in the Cu/Sn/Cu bonding joint. On the other hand, when using the Ar plasma surface-treated Sn preform, IMCs formed throughout the entire bonding joint. It is known that Ni₃Sn₄ IMCs are formed in the Ni-Sn system¹³⁾, and Ni-Sn-P and Ni₃P are also known to form at the interface between Sn-based solders and electroless Ni(P)¹⁷^-²⁰⁾. Similar to a previous research, in this study, Ni-Sn-P and Ni₃P were also observed at the interface between the Ni(P) surface-finished DBC substrate and Sn preform. Meanwhile, as the bonding time increased to 30 and 60 minutes, Ni₃Sn₄ IMCs formed throughout the entire bonding joint even without Ar plasma surface treatment. Fig. 7 shows the ratio of IMCs formed at the bonding joint for various conditions of Ar plasma surface treatment and bonding time. In the case of 10 minutes of bonding time without Ar plasma surface treatment (Ni-10), Ni₃Sn₄ IMCs had the lowest ratio at the bonding joint due to the influence of residual Sn that could not form IMCs at the bonding joint. However, as the bonding time increased to 30 and 60 minutes (Ni-30 and Ni-60), the residual Sn reacted and ratios of Ni₃Sn₄, Ni-Sn-P, and Ni₃P were increased. While Ni-Sn-P and Ni₃P partially increased with the bonding time after Ni₃Sn₄ IMCs formed throughout the entire bonding joint, the ratios of Ni₃Sn₄, Ni-Sn-P, and Ni₃P were not remarkably increased. Moreover, in the Ar plasma-treated bonding joints, Ni₃Sn₄ IMCs formed throughout the entire bonding joint from a bonding time of 10 minutes (Ni(Ar)-10) and the ratios of IMCs forming the bonding joint were not noticeably changed even as the bonding time increased to 30 and 60 minutes (Ni(Ar)-30, Ni(Ar)-60).

Fig. 6

Cross-sectional SEM micrographs of Ni/Sn/Ni joints with various bonding times

Fig. 7

Ratio of IMC formed at Ni/Sn/Ni joints with (Ar(Ni)) and without(Ni) plasma surface treatment

The bonding strengths of the Ni/Sn/Ni TLP bonding joint were showed in Fig. 8 and these results were a similar trend to the bonding strength results of the Cu/ Sn/Cu TLP bonding joint. For the bonding joint using an Ar plasma surface-treated Sn preform at a bonding time of 10 minutes (Ni-10), the bonding joint showed a high strength of 43 MPa due to the residual Sn. It was assumed that this was attributed to the ductility characteristics of the residual Sn, as described for the Cu/ Sn/Cu TLP bonding joint. However, as the bonding time increased to 30 and 60 minutes (Ni-30, Ni-60), the bonding strength was about 23-25 MPa after Ni₃Sn₄ formed throughout the entire bonding joint and the strength was not greatly changed with increasing the bonding time. Meanwhile, as in the case of Ar plasma surface treatment, Ni₃Sn₄ IMCs formed throughout the entire bonding joint from the bonding time of 10 minutes (Ni(Ar)-10) and the bonding strengths were similar with the initial strength at 10 minutes although the bonding time increased to 30 and 60 minutes (Ni(Ar)- 30, Ni(Ar)-60). This means that the Ar plasma surface treatment and the bonding time could not remarkably affect to the bonding strength when Ni₃Sn₄ IMCs form throughout the entire bonding joint in the Ni/Sn/Ni structure. Fig. 9 shows the cross-sectional analysis results of the fracture loci of the Ni/Sn/Ni TLP bonding joint and a schematic diagram of fracture occurrence with various conditions of bonding time and plasma surface treatment. Similar to the Cu/Sn/Cu TLP bonding joint, the bonding joint fractures occurred mostly at the substrate-side of the bonding joint. When residual Sn remains at the bonding joint without plasma treatment at the bonding time of 10 minutes, as shown in Fig. 9 (a) and (b) Case 1, the fracture loci occurs at the interface between Ni₃Sn₄ IMC and residual Sn. When the bonding time increased to 30 and 60 minutes, however, fractures occurred at the Ni₃Sn₄ IMC, Ni₃Sn₄, and Ni-Sn-P interfaces due to the formation of Ni₃Sn₄ IMCs throughout the whole joint. In contrast, when Ar plasma treatment was applied as shown in Fig. 9 (a) and (b) Case 2, fractures occurred at the Ni₃Sn₄ IMC, Ni₃Sn₄ and Ni-Sn-P interfaces from a bonding time of 10 minutes. Due to the similarity of the fracture mode, bonding strengths in the bonding joint of 30 minutes bonding time without plasma treatment (Ni-30) and those after 10 minutes bonding time with Ar plasma treatment (Ar(Ni)-10) were similarly measured.

Fig. 8

Shear strength of Ni/Sn/Ni joints for various bon- ding times and Ar plasma treatment

Fig. 9

(a) Cross sectional SEM micrographs of fracture surface and (b) fracture mode of Ni/Sn/Ni joints after shear test

4. Conclusions

As a method to shorten the TLP bonding process time for power modules, we applied Ar plasma treatment to the surface of the Sn preform insert metal used as the bonding material and investigated the effect of Ar plasma treatment on the mechanical properties and interfacial reactions of the TLP bonding joint. In addition, TLP bonding joints with Cu/Sn/Cu and Ni/Sn/Ni structures were formed and investigated the properties of bonding joints with various TLP bonding conditions for Ar plasma treatment, TLP bonding joint structures, and TLP bonding time.

1) When the Sn preform was applied to Ar plasma treatment, even at a bonding time of 10 minutes, IMCs formed throughout the entire Cu/Sn/Cu and Ni/Sn/Ni bonding joints. It was assumed that the diffusion of metal atoms such as Cu, Ni, and Sn was influenced due to the physical cleaning effect and so on by the Ar plasma treatment.
2) When Ar plasma treatment was not applied, at the bonding time of 10 minutes, IMCs were not fully formed throughout the entire bonding joints of Cu/Sn/Cu and Ni/Sn/Ni structures and residual Sn was observed.
3) Cu₆Sn₅ and Cu₃Sn IMCs formed in the Cu/Sn/Cu TLP bonding joint while Ni₃Sn₄, Ni-Sn-P, and Ni₃P IMCs were observed in the Ni/Sn/Ni TLP bonding joint.
4) The ratio of Cu₃Sn IMCs in the Cu/Sn/Cu bonding joint increased as the bonding time increased and this affected to increase the strength of the bonding joint.
5) In the Ni/Sn/Ni bonding joint, the Ar plasma surface treatment and the bonding time could not remarkably affect to the bonding strength when Ni₃Sn₄ IMCs were fully formed throughout the entire bonding joint.
6) From these results, when using a preform insert metal as the bonding material for TLP bonding, Ar treatment on the preform is expected to shorten the process time by quickly forming the entire TLP bonding joint through IMCs.