1. Introduction
The high-output power module market is expanding with the increasing demand for electric vehicles. A power module, a key component of an electric vehicle, consists of a semiconductor device, interconnection, a substrate, a base plate, a heat sink, and an epoxy molding compound (EMC). Since the temperature of the power module increases during operation alongside the increase in its output, the reliability of the device joints at high temperature is very important and the mitigation of thermal stress is required1-5).
The ceramic substrate structure for the power module is the Cu/ceramic/Cu structure, and the ceramic materials mainly used are aluminium oxide (Al2O3), aluminium nitride (AlN), and silicon nitride (Si3N4). The substrate types are distinguished by the method of bonding ceramics and Cu. The representative types include the direct bonding copper (DBC) substrate, which directly bonds ceramics and Cu, and the active metal brazing (AMB) substrate, which applies brazing filler metal (BFM) to improve the wettability of the bonding interface. The DBC substrate is bonded without applying BFM mainly through oxidation reduction reactions with Al2O3 and Cu. In the case of AlN with no Al2O3 layer, the surface is oxidized as preprocessing to form the Al2O3 layer. In the 1,065-1,085°C temperature range, Cu then reacts with oxygen on the ceramic surface to form a eutectic liquid of Cu-O, when cooled down, the Cu-Cu2O eutectic solution solidifies to form a joint6). In the case of the AMB substrate, the brazing method that utilizes the BFM layer does not require preprocessing for surface oxidation, and bonding is possible within a relatively low temperature range between 800 and 900°C.
It has been reported that the AMB substrate has higher mechanical reliability than the DBC substrate by providing the continuity of joints through the metallurgical combination of the BFM layer and ceramics7-9). BFM types include Cu-Ti, Cu-Zr, and Ag-Ti systems that added reactive elements, such as Ti, Zr, and Cr, to brazing metals that are commonly used as solvent metals, such as Cu, Ag, and Sn. Also, various ternary and quaternary alloys have been developed by adding metals to improve low-temperature processes and reliability8-11). Particularly, wettability with ceramics significantly improves as the Ti content increases to approximately 5 at.%, and the presence of Ag in the BFM/ceramic reaction is known to further activate the interface reaction of Ti12).
Since the BFM layer in the AMB substrate serves as the movement path of heat, increasing thermal resistance, it is very important to adjust its thickness13). The BFM process is mainly applied in the form of a thin film, powder, or paste. Therefore, there are limitations in reducing the thickness of the BFM layer.
In this study, the nano-sputtering process was applied to minimize the BFM layer thickness of AlN and Si3N4 AMB substrates. In addition, the Ag-Cu-Ti BFM layer was deposited by adopting Ag-Cu solvent alloys that can reduce the process temperature compared to Cu and Ag solvent metals and Ti reactive element of excellent reactivity with nitrides. As for the composition of BFM, Ti 5 at.% that improves wettability with ceramics and Ag 71.9-91.2 wt.% in the Ag-Cu solvent alloys were judged to be appropriate, which is the melting reaction section of Ag that increases the Ti reactivity. After the deposition of the BFM layer, a Cu thin film was bonded using the brazing process. Silicon carbide (SiC) devices of the power module were bonded through Sn-3.0Ag- 0.5Cu (SAC305) soldering and Ag sintering processes, and bonding characteristics before and after a thermal shock test were compared and analyzed.
2. Experimental Method
2.1 Ag-Cu-Ti nano sputtering
For the production of the AMB substrate, 0.38㎛-thick AlN, 0.32㎛-thick Si3N4, and 0.3㎛-thick Cu thin film were used. The nano-sputtering process was performed 70 and 100 minutes at 8.3 kW, respectively, using Ag, Cu, and Ti target. The surface microstructure and thickness over time were analyzed. A scanning electron microscope (SEM) and an energy dispersive spectrometer (EDS) were used for microstructure analysis. The deposition thickness was measured after applying focused ion beam (FIB) milling to the cross section.
2.2 Ceramic/Cu brazing bonding
The ceramics with the BFM layer and the Cu thin film were subjected to 1h brazing bonding at 850°C under a vacuum atmosphere. For bonding strength measurements, peel strength test samples were prepared for each ceramic type shown in Fig. 1(a, b), and peel strength was measured using the jig as shown in Fig. 1(c, d). In addition, BFM layers for each ceramic type were compared through cross-sectional microstructure analysis.
2.3 SiC device solder and sinter joints
The SiC device used in the experiment was a shottky diode in a size of 3.15×3.15 mm. Type 4 SAC305 solder paste (Tamura Chemical Co., Ltd.) and Ag film (ALPHA) were used as bonding materials. Fig. 2 shows the pattern of the AMB substrate.
For SAC305 solder paste bonding, the vacuum reflow process was applied with the temperature profile shown in Fig. 3(a). Ag film bonding was performed on the Cu surface of the AMB substrate by applying the pressure sintering process with the temperature profile shown in Fig. 3(b). After the processes, bonding characteristics according to each process were compared through shear strength measurement and SEM/EDS cross-sectional microstructure analysis.
2.4 Thermal shock test
The ceramic substrates and SiC device modules were subjected to the thermal shock test in accordance with the related standards. The thermal shock test of the AMB substrate was conducted by 500 cycles under -55-150°C and 15 min dwell conditions in accordance with the JESD22-A04E standard. The thermal shock test of the SiC device-bonded samples was conducted by 500 cycles under -40-125°C and 15 min dwell conditions in accordance with the AQG 324, 8.2 QE-01 standard. The deterioration characteristics of the ceramic/Cu joint and SiC device/Cu joint were compared and analyzed through bonding strength measurement and cross-sectional microstructure analysis before and after the test.
3. Experiment Results
3.1 BFM (Brazing Filler Metal) surface microstructure analysis
The thickness of Ag-Cu-Ti BFM after nano sputtering was approximately 5 ㎛ for the 70 min sample and 7 ㎛ for the 100 min sample. When the surface microstructure was observed, it was observed that the surface was flatter and the grain size was larger for the 100 min sample compared to the 70 min sample as shown in Fig. 4. While all of the Ag-Cu-Ti components were detected in the flat area (1) of Fig. 4(a), only the Ag-Cu components were observed in the non-dispersed area (2) (Table 1). When a large area was subjected to EDS analysis at a low magnification, the surface composition of BFM was measured to be Ag-15.0Cu-2.3Ti.
3.2 Analysis of brazing bonding characteristics
When the cross-sectional microstructure of the brazing joints of the AMB substrate was analyzed, it was observed that the BFM layer was formed as thin as approximately 10 ㎛ as shown in Fig. 5 (a, d). In the BFM layer, Cu areas were distributed in the form of spheres in Ag areas as shown in Fig. 5 (b, e). In the formed reaction layers, the Ti, Cu, and Ag components were detected at the AlN/BFM interface as shown in Fig. 5(c) while the Si, Ti, Cu, and Ag components were observed at the Si3N4/BFM interface as shown in Fig. 5(d). According to previous studies, TiN compounds were formed at the AlN/BFM interface14) while Ti5Si3 intermetallic compounds (IMCs) and TiN compounds were formed at the Si3N4/BFM interface15) during the brazing process with Ag-Cu-Ti BFM. Therefore, area 3 in Fig. 5(b) is expected to be TiN compounds. The boundary of the reaction layer was observed very clearly. Areas 4 and 5 in Fig. 5(e) are expected to be Ti5Si3 IMCs and TiN compounds. As they were formed in various shapes, such as layers and spheres, the boundary of the reaction layer was not clear compared to AlN. The reaction formula of Ti and Si3N4 is [9Ti+ Si3N4 = 4TiN+Ti5Si3], and Gibbs free energy (△G) is -1170.647 kJ/mol, indicating the formation of TiN compounds and Ti5Si3 IMCs through voluntary reactions. The Gibbs free energies of TiN and Ti5Si3 are -346.890 and -644.059 kJ/mol, indicating that Ti5Si3 IMCs are easier to form16). Therefore, it is speculated that there are more Ti5Si3 IMCs than TiN compounds in the compounds with various shapes formed at the Si3N4/BFM interface.
Fig. 5
Cross-sectional (a,b,d,e) SEM micrographs and (c,f) EDS analysis results of (a-c) AlN and (d-f) Si3N4 AMB substrates
Fig. 6 shows the results of measuring the bonding strength of the ceramic/Cu interface. As shown in Fig. 6(a), peel strength of the ceramic/Cu interface was higher in the Si3N4/Cu sample than the AlN/Cu sample. In the case of the Si3N4/Cu sample, the ceramic layer was fractured instead of the bonding interface as shown in Fig. 6(c). Therefore, the actual bonding strength of the Si3N4/Cu sample was considered higher than the measured value. When the fracture surface of the AlN/ Cu sample was analyzed, ceramic components were observed in area 1 and BFM components in area 2 as shown in Fig. 6(d-e), confirming that the fracture occurred at the ceramic/BFM interface. The Ti5Si3 IMCs formed at the Si3N4/BFM bonding interface are known to have high strength, chemical stability, acid resistance, and mechanical properties17). In the case of TiN compounds, brittleness tends to increase as the N content exceeds 0.9 wt.%18). Therefore, it is determined that the type and condition of the IMC layer at the bonding interface formed by the diffusion reaction of the ceramic layer and BFM elements affected the metal-ceramic bonding strength.
3.3 Analysis of solder and sinter joint microstructure
Fig. 7 shows the cross-sectional microstructure analysis results after SAC305 soldering and Ag film sintering bonding to the AMB substrate. There was no difference depending on the ceramic material because the Cu materials on the surface of bonded substrate are the same. As shown in Fig. 7(a-b), the thickness of the Ag sinter joint was much lower than that of the solder joint. Cu6Sn5 IMCs were observed at the solder joint interface as shown in Fig. 7(d) while excellent bonding with dense Ag particles and no large void was observed at the sinter joint as shown in Fig. 7(f).
3.4 Analysis of brazing joint microstructure after the thermal shock test
Fig. 8 shows microstructure analysis results for the brazing joints of SiC device sinter joint samples before and after the -40-125°C thermal shock test. For AlN, very large spherical Cu areas were observed inside the Ag matrix of the BFM layer as shown in Fig. 8(a-b). In the case of Si3N4, Cu areas in the form of dense lamellae were observed as shown in Fig. 8(d-e). This difference in the geometry of Cu areas depends on the growth rate of Cu. It was reported that the lamella structure is observed when the growth rate is low and the spherical form when it is high19). The thermal conductivities of AlN and Si3N4 are 180 and 90 W/mK, respectively, and it was judged that the growth rate of Cu was higher at the AlN interface with higher thermal conductivity. After the thermal shock test at the AlN/BFM interface, Ti was detected in area 4 inside AlN as shown in Fig. 8(b-c). This appears to be due to the diffusion of Ti in the thermal shock test. At the Si3N4/BFM interface, on the other hand, the Ti content increased in area 7 in the reaction layer as shown in Fig. 8 (e-f), but Ti was not detected in area 8 inside Si3N4.
Fig. 8
Cross-sectional (a,b,d,e) SEM micrograph and (c,f) EDS analysis results of (a-c) AlN/BFM and (d-f) Si3N4/BFM interface, (a,d) as-bonded, (b,e) after thermal shock test
After the -55-150°C thermal shock test of the AMB substrate, delamination was observed from the AlN substrate as shown in Fig. 9(a). The cross-sectional microstructure analysis results for the delaminated section showed that cracks started at the AlN/BFM interface and progressed into the inside of the ceramic layer as shown in Fig. 10(a). These results showed similar tendencies to the fracture surface analysis results after the peel strength test shown in Fig. 6(d-e). According to a previous study, the TiN layer grows along the AlN grain boundary and becomes thicker until TiN is completely formed on the AlN surface, and the porous TiN layer becomes a weak link and reduces the bonding strength of the interface. This porous TiN layer is formed when the TiN/AlN volume ratio is lower than 1, and it is known to form a molten BFM channel along the AlN grain boundary due to its contraction tendency compared to AlN20). In this study, the TiN/AlN volume ratio was much lower than 1, and BFM components that penetrated into the ceramics were observed as shown in Fig. 10(a). Therefore, it was determined that the fracture was caused by the thermal shock at the AlN/BFM interface weakened by the formation of the porous TiN layer and BFM channel.
4. Conclusion
In this study, the nano-sputtering process of Ag-Cu-Ti brazing filler metal (BFM) was applied to AlN and Si3N4 active metal brazing (AMB) substrates to improve the ceramic/Cu bonding characteristics of substrates for silicon carbide (SiC)-based power modules, and bonding characteristics were compared through peel strength and thermal shock tests. In addition, the applicability of the SiC device for power modules was examined through SAC305 solder and Ag film sinter joints.
1) For the nano-sputtering thin film of Ag-Cu-Ti BFM, the surface roughness decreased under the 100 min condition compared to the 70 min condition, and the surface composition of the entire area was observed to be Ag-15.0Cu-2.3Ti.
2) After AMB bonding, the average thickness of the BFM layer was observed to be as low as approximately 10 ㎛. The peel strength of the ceramic/Cu interface was measured to be 2.35 kgf/mm for AlN and 4.26 kgf/mm for Si3N4. TiN compounds were observed at the AlN/BFM interface and the reaction layer estimated to be Ti5Si3 intermetallic compounds (IMCs) and TiN compounds at the Si3N4/BFM interface. This reaction layer appears to have affected the metal-ceramic bonding strength.
3) When cross-sectional microstructure analysis was conducted after SAC305 soldering and pressure Ag sintering bonding processes, it was observed that the Ag pressure sinter joint formed a very low thickness, and that the SiC device was well bonded without defects on both AlN and Si3N4 AMB substrates.
4) After the thermal shock test of 500 cycles under -55-150°C and 15 min dwell conditions, no defect was observed from the Si3N4 AMB substrate, but delamination was observed from the AlN substrate. Cross-sectional analysis confirmed that the fracture occurred in the TiN reaction layer of the AlN/BFM interface.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Print
