1. Introduction
In recent years, the use of SiC- and GaN-based wide-bandgap (WBG) semiconductor devices has expanded rapidly in the packaging field1,2). As operating conditions move toward higher temperatures and higher power levels, securing a reliable die-attach technology has become a critical requirement. As power density increases, the heat generated in the chip must be efficiently transferred to the substrate, and the thermal and mechanical stability of the joint directly affects the long-term reliability of the module3-6).
High-thermal-conductivity ceramic substrates such as aluminum nitride (AlN) and silicon nitride (Si3N4) are widely used in power devices, heaters, and heat-spreading modules that operate under harsh thermal environments. These ceramics offer a high thermal conductivity of approximately 170-200 W/m·K and a low coefficient of thermal expansion (CTE), enabling fast heat dissipation while maintaining excellent electrical insulation and mechanical strength. Owing to these characteristics, the structural integrity of the substrate can be maintained even under high-power operation or repeated temperature cycling5,7,8).
Although AlN and Si3N4 substrates can withstand elevated temperatures above 250 °C, such environments accelerate thermal and mechanical degradation within the die-attach layer4,9). Conventional solder joints (Sn-Ag-Cu, Sn-Pb, etc.) suffer from low melting points, recrystallization during thermal cycling, intermetallic compound (IMC) growth, and accumulated thermal fatigue or creep deformation. These degradation mechanisms lead to cracking, interfacial delamination, and increased thermal resistance, ultimately limiting long-term reliability10).
For these reasons, alternative die-attach materials capable of maintaining structural stability and high thermal conductivity at elevated temperatures are increasingly required. Ag-sintered joining, a solid-state diffusion-based process without a melting phase, provides excellent high-temperature stability and mechanical robustness. Additionally, the high thermal conductivity of Ag makes it particularly suitable for ceramic-based power modules7,11,12).
The reliability of Ag-sintered joints is strongly influenced not only by material and process conditions but also by factors such as bond-line thickness (BLT), porosity, and interfacial condition4,9,13). Sihai Chen et al. reported that pressure-less sintering achieves void fractions below 2% when the BLT is maintained at 40-60 μm, resulting in stable shear strength even after thermal cycling. In contrast, excessively thin joints exhibited interfacial degradation due to insufficient diffusion, whereas thick joints showed increased thermal resistance caused by internal void formation11). These findings indicate that BLT plays a key role in balancing thermal performance and mechanical stability. Although the thickness effect on Ag-sintered joints has been reported experimentally, simulation-based quantitative evaluations of thermal stress distribution in ceramic substrate structures remain limited.
In this study, Ag-sintered joints with various BLTs were fabricated, and their mechanical properties and failure behavior were evaluated after thermal shock testing. Finite element analysis (FEA) was then used to quantify the differences in thermal stress and deformation behavior according to BLT, and the results were compared with experimental observations. This combined approach provides insight into how BLT influences the thermal-shock reliability of Ag-sintered joints and suggests design guidelines for achieving improved thermal and mechanical stability.
2. Experimental prdcedure
2.1 Materials and components
The substrate used for the sintered joints was a 4 mm thick AlN wafer with a 12-inch diameter, which was diced into pieces measuring 15 × 8 mm2. A metallization pad with a diameter of 7 mm was formed on the AlN surface, and an electroless nickel immersion gold (ENIG) finish was applied. The power terminals were fabricated using Ni with a diameter of 5 mm and a height of 2.8 mm, and the same ENIG metallization was used. The geometry and configuration of the substrate and terminals are shown in Fig. 1(a). A commercial Ag sintering paste (MAX102, Nihon Handa, Japan) was used as the attach material. The Ag powder in the paste consisted of spherical particles with an average size of 322±75 nm, as illustrated in Fig. 1(b).
2.2 Preparation of samples
Ag paste was printed onto the ENIG-metallized pads of the AlN substrates using stencil masks with thicknesses of 50, 100, and 200 μm. After printing, the Ni power terminal was positioned on top of the paste, and bonding was carried out using a thermo-compression bonding system (SJCWD-100, SJ Company, Korea). A schematic of the thermo-compression process is presented in Fig. 2(a), and the sintering procedure followed the temperature-pressure profile shown in Fig. 2(b). The samples were first pre-heated at 100 °C for 10 min to remove residual solvent, and then sintered at 280 °C for 30 min under an applied pressure of 10 MPa. All processes were performed under ambient conditions. After sintering, the specimens were evaluated through shear test, thermal-shock test, and microstructural characterization.
2.3 Evaluation Mechanical and microstructural properties
To evaluate the mechanical properties of the Ag-sintered joints, shear test was conducted using a shear tester (Dage 4000, Dage, USA). The shear condition was set to a shear speed of 200 μm/s and a shear height of 200 μm. After thermal-shock test, the shear strength data were analyzed to examine the correlation with the bond-line thickness conditions (50, 100, and 200 μm). Microstructural and interfacial observations of the Ag-sintered joints were carried out using a field-emission scanning electron microscope (FE-SEM; Inspect F, FEI Co., USA). The evolution of the joint microstructure was examined for each bond-line thickness after 0, 500, and 1,000 thermal shock cycles. Thermal shock test was assessed following the temperature profile shown in Fig. 3, cycling between -45 °C and 150 °C. Each temperature step was maintained for 15 minutes, and a total of 1,000 cycles was performed. The same thermal shock conditions were applied to the simulation model to compare the thermal stress behavior with the experimental results.
2.4 Simulation method
Finite element analysis was performed using ANSYS to conduct a static structural analysis and calculate the stress and strain distributions. The three-dimensional analysis model used in the simulation is shown in Fig. 4. The model consisted of the AlN substrate, the Ag-sintered joint, and the Ni power terminal, and the geometry and dimensions were matched to those of the actual specimens (AlN: 15 × 8 mm, Ni: Ø5 mm). The BLT was defined by incorporating the actual shrinkage measured after the bonding process so that the simulation could closely represent the experimental conditions. Material properties were assigned based on values reported in previous studies, and the key properties are summarized in Table 1. As shown in Fig. 4, the mesh was generated for the three-dimensional model, and the bottom surface of the AlN substrate was fixed, while the top surfaces were assigned free boundary conditions. The thermal shock temperature profile used in the reliability test was applied identically to the simulation. This allowed the change in thermal stress distribution with respect to BLT to be evaluated, and the results are discussed in the following section.
3. Results and Discussion
3.1 Microstructural evolution with different bond-line thicknesses
Fig. 5 shows the cross-sectional microstructures of the Ag-sintered joints fabricated with different printing thicknesses. When the initial printing thicknesses were set to 50, 100, and 200 μm and sintered under identical conditions (250 °C, 30 min, 5 MPa), the resulting BLTs were measured to be approximately 33 μm, 67 μm, and 142 μm, respectively. As summarized in Fig. 6, these values correspond to a shrinkage of about 25-35% relative to the initial printed thickness. The BLT measurements indicate that thicker printed layers exhibited lower shrinkage after sintering under the same processing conditions. The reduction in BLT is attributed to solvent evaporation, particle compaction under applied pressure, and diffusion-driven densification during sintering. When the printed layer is thin, solvent escapes rapidly between the paste surface and the substrate, and particle rearrangement under pressure becomes more concentrated, resulting in a higher shrinkage ratio. In contrast, thicker printed layers restrict internal solvent evaporation and reduce localized stress generated during pressing, leading to lower shrinkage. This behavior is consistent with previous observations indicating that the residual porosity of Ag-sintered joints changes with BLT and applied pressure, and that thicker printed layer exhibit more gradual shrinkage during sintering4,9).
3.2 Failure behavior
Fig. 7 compares the cross-sectional microstructures of specimens with different BLTs (50, 100, and 200 μm) before and after the thermal shock test. SEM images were taken at 0, 500, and 1,000 cycles, and each specimen was observed at the same location for consistent comparison. In the initial state before the thermal shock test, all specimens exhibited uniformly developed necks between Ag particles, and micro-voids were evenly distributed throughout the joint. No cracks or reaction layers were observed at the interfaces between the Ag-sintered joint, the substrate, and the Au metallization layer of the power terminal. After 500 cycles, the specimen with a BLT of 50 μm, exhibited localized cracking near the interfaces between the Ag-sintered joint and both the substrate and the terminal. The cracks propagated along the neck boundaries between Ag particles and, in some regions, extended toward the interface. After 1,000 cycles, continuous interfacial delamination developed along the joint, which is attributed to the relatively large stress concentration near the interface caused by the CTE mismatch and the thin BLT4). In the 100 µm BLT specimen, no significant cracks or delamination were observed after the thermal shock test, and the micro-void structure remained substantially unchanged. However, cracks formed near the edge of the power terminal after 1,000 cycles. This localized failure is likely due to stress concentration at the specimen edge caused by repeated thermal expansion. For the 200 μm BLT specimen, no appreciable structural changes were observed within the Ag-sintered joint after thermal shock. Although small cracks appeared near the substrate interface after 1,000 cycles, their propagation was limited, and the continuity of the joint layer was maintained. Fig. 8 shows crack behavior for each BLT after 1,000 cycles. The 50 μm joint exhibited continuous crack propagation along Ag neck boundaries near the interface due to high stress concentration. In the 100 μm joint, cracks formed locally at the terminal edge and propagated toward the center, while the overall microstructure remained more stable than in the 50 μm joint. In the 200 μm joint, cracking was limited and appeared mainly at the edge interface between the substrate’s ENIG layer and the Ag neck region, indicating that crack initiation can occur at the peripheral interface. Consequently, this behavior suggests that a thicker BLT more effectively accommodates thermal expansion within the joint, reducing the stress transferred to the interface4,14). Overall, the fracture behavior under thermal shock varied with BLT. The thinnest joint experienced localized cracking due to high stress concentration, whereas the thicker joints distributed stress more uniformly, resulting in reduced deformation and improved stability.
3.3 Shear strength
Fig. 9 shows the change in shear strength for different BLTs as a function of thermal shock cycles. The initial shear strengths of the 50, 100, and 200 μm specimens were 29.6, 33.4, and 33.0 MPa, respectively, indicating that all three conditions exhibited comparable initial strength. After 500 cycles, the strengths decreased to 8.8, 22.6, and 31.8 MPa, and further dropped to 3.1, 18.0, and 24.4 MPa after 1,000 cycles. The specimen with a BLT of 50 μm exhibited the most severe degradation, showing a significant decrease in shear strength with repeated thermal shock. This behavior is attributed to early crack initiation and fast crack propagation near the interface where thermal stress is highly concentrated. The 100 μm specimen initially showed the highest shear strength and maintained a level above 22 MPa even after 500 cycles. Although the strength decreased to 18 MPa after 1,000 cycles, the reduction was more moderate because the BLT was sufficient to distribute thermal stress and delay interfacial damage and crack growth. For the 200 μm specimen, the shear strength remained relatively high after the thermal shock test, decreasing to 31.8 MPa after 500 cycles and 24.4 MPa after 1,000 cycles, representing the smallest drop among all conditions. Despite accumulated thermal shock cycles, interfacial fracture was limited, and microcracks were observed mainly near the edge of the substrate interface. Overall, although all specimens experienced a reduction in shear strength with repeated thermal shock, the rate of degradation depended on the BLT. The thinner joint showed a sharp decrease in strength, whereas BLTs of 100 μm or greater effectively mitigated stress concentration and provided improved mechanical stability under thermal shock.
3.4 Finite Element Analysis
The finite element analysis was performed using a static structural analysis, and the calculations were carried out based on the governing equation shown in Eq. (1).
Here, F represents the applied force, u is the displacement, and K denotes the stiffness matrix. The Ni-Ag-AlN layered structure was modeled, and the thermal shock conditions were simulated for BLTs of 50, 100, and 200 μm. To reflect the actual shrinkage after sintering, the BLTs were set to 30, 70, and 140 μm in the analysis. The stress response of the Ag-sintered joint was evaluated using the von Mises criterion, and the equivalent stress and equivalent strain were calculated accordingly. The equivalent stress was obtained using the Eq. 2 below15-17).
In this equation, σy, σy, and σz denote the normal stresses along each principal direction, while τxy, τyz, and τzx represent the shear stresses acting on the corresponding planes. These stress components are combined into a single equivalent stress value to describe the total stress within the structure. Similarly, the equivalent strain is calculated using both normal and shear strain components, as expressed in Eq. (3).
Here, εx, εy, and εz represent the normal strains in each direction, and γxy, γyz, and γzx denote the shear strains. These components were combined into a single equivalent strain value to quantify the overall deformation under thermal shock. Fig. 10(a-c) shows the equivalent stress distributions at 150 °C for each BLT. In Fig. 10(a), the 50 μm joint shows strong stress concentration along the edges of the Ni/Ag and Ag/AlN interfaces. This occurs because repeated thermal expansion and contraction create large shear stress in the thinner joint due to the CTE mismatch. As the BLT increases to 100 and 200 μm, the stress becomes more evenly distributed, and the high-stress regions become smaller. Fig. 11 summarizes the maximum equivalent stress for each BLT as the thermal shock cycles increase. When the BLT increased from 50 to 100 to 200 μm, the maximum stress decreased by about 25% and 44%, confirming the stress-relief effect of a thicker joint. The equivalent strain distributions in Fig. 12 shows a similar trend. High-strain regions appeared near the interface edges, and thinner joints showed higher strain.
Fig. 10
Cross-sectional equivalent stress distributions of Ag-sintered joints with different bond-line thicknesses, (a) 50 µm, (b) 100 µm, and (c) 200 µm
Fig. 11
Variation of the maximum equivalent stress of Ag-sintered joints during thermal shock. Cycles
Fig. 12
Equivalent strain distributions of Ag-sintered joints with different bond-line thicknesses (50, 100, and 200 µm)
The crack behavior observed in the SEM images matched the simulation results. For the 50 μm joint, cracks and partial delamination were found near the Ni/Ag and Ag/AlN interface edges after thermal shock, which corresponds to the high-stress and high-strain regions predicted by the FEA. In contrast, the 100 and 200 μm joints showed almost no interfacial damage, and the FEM results also showed more uniform stress distribution in these thicker joints. Overall, increasing the BLT reduced stress concentration during thermal shock and improved interfacial stability.
4. Conclusions
In this study, the effect of the Ag-sintered joint thickness on thermal shock reliability was examined through experiments and finite element analysis. Based on the SEM observations, shear test results, and Finite Element Analysis, the following conclusions were drawn.
1) For printing thicknesses of 50, 100, and 200 μm, the actual BLTs after sintering were approximately 33, 67, and 142 μm, respectively. Thicker joints showed lower shrinkage during sintering.
2) The FEA results showed that the equivalent stress decreased from 251 MPa to 141 MPa (about 44%) as the BLT increased from 50 to 200 μm. High stress and strain were concentrated near the interface edge for the 50 μm joint.
3) Stress concentration caused by the CTE mismatch was reduced as the BLT increased, resulting in a more uniform stress distribution.
4) The crack locations observed in the SEM images agreed well with the high-stress regions predicted by the FEA, confirming that BLT affects both thermal and mechanical reliability.
Overall, optimizing the BLT is an important design factor for improving the reliability of Ag-sintered joints used in high-temperature and high-power packaging applications. Nevertheless, not all geometry related factors such as fillet and morphological variations in joint were fully considered in this study, and these aspects may require further examination under more carefully controlled conditions.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Print
