Evolution of curvature under thermal cycling in sandwich assembly bonded by sintered nanosilver paste

Abstract

PurposeThe purpose of this paper is to determine: how much the residual curvature could be formed in sintered nano‐silver assembly when it is cooled to room temperature from the sintering temperature (normally 275°C); how the cyclic temperature load affects the residual curvature or stresses in sintered joint. Then the stress level and the reliability of sintered nano‐silver for high‐temperature applications can be understood.Design/methodology/approach5 mm * 2.5 mm silicon chip was bonded with 96 per cent Al2O3 substrate by sintering nanosilver paste. An optical system was developed to measure the curvature of the sintered assemblies. Reliability of the sintered assemblies was evaluated by temperature cycling of −40∼125°C. Finite element analysis was employed to simulate the behavior of the joint subjected to the temperature cycling from −40°C to 125°C by ANSYS. SEM images were taken to investigate the impact of temperature cycling on the reliability of sintered silver attachment.FindingsThis residual bending at room temperature was found concave towards the substrate (alumina) side. Also, with the bondline thickness increasing, the residual curvature decreases obviously. The severity of the residual bending in all the structures was mitigated to some extent with increasing number of cycles. There is no crack in the joint with the thickness of 25 μm. The drop of the residual curvature of the samples with bondline of 25 μm is caused mainly by stress relaxation in sintered silver before 300 cycles. Sample with thicker bondline is more susceptible to thermal cycling for the structure bonded with nanosilver than that with thinner bondline. The poor quality of bonding is due to the thicker sintered joint, which means that sintered nanosilver is not suitable for die‐attachment requiring thick bondline.Originality/valueThe paper describes: how a precise optical system was developed to measure the residual curvature of the sintered assemblies; how the evolution of the residual curvature of the sintered assembly with the temperature cycling was obtained by both experiment and simulation; and how microstructures of the sintered silver joint were analyzed for as‐sintered assembly and the sintered assembly after temperature cycling.

Are Sintered Silver Joints Ready for Use as Interconnect Material in Microelectronic Packaging?

“…71 This constraint on the die-attach thickness differs from the findings of an earlier report that recommends a thickness of more than 100 lm to produce a fatigue-resistant joint between a DCB and a copper base plate. 62 This difference can be attributed to the different bonding areas: an area of 5 mm 9 2.5 mm was used in the former study, 71 while the latter study used a bonding area of 23 mm 9 18 mm. 62 Hence, caution must be exercised when adopting process parameters from different bonding interfaces in power electronics for copper lead frame products used in other sectors.…”

“…63,64,69,71,87,88 These lifetime-prediction models complement thermalfatigue lifetime studies, as mentioned earlier. Two of the more successful models that have been investigated for sintered Ag joints are the combined Navarro-de los Rios and small cracks interaction physical model 69 and the modified Basquin model.…”

“…71 According to that study, sintered Ag joints were found to be unsuitable for die-attach thicknesses of 75 lm or more because of their increased propensity for crack growth instead of stress relaxation within the sintered Ag layer. 71 This constraint on the die-attach thickness differs from the findings of an earlier report that recommends a thickness of more than 100 lm to produce a fatigue-resistant joint between a DCB and a copper base plate. 62 This difference can be attributed to the different bonding areas: an area of 5 mm 9 2.5 mm was used in the former study, 71 while the latter study used a bonding area of 23 mm 9 18 mm.…”

“…[62][63][64] Some researchers have defined a crack length equal to the thickness of the die-attach layer as the detection limit for observing crack lengths and failure criteria. 71 Therefore, caution should be exercised when comparing results from different sources. The results of various thermal-cycling tests regarding the fatigue resistance (i.e., cycle to failure, N f ) of sintered Ag joints are shown in Table III.…”

“…71 There are inherent limitations in this model, which is not capable of handling the evolving microstructures of these sintered Ag joints. Other lifetime-prediction models have also been borrowed from studies conducted using Pb-Sn or Pb-free solder joints.…”