Test Methods for Silicon Die Strength

Recently, the 3D or stacked-die packages become increasingly popular for packaging ICs into a system or subsystem to satisfy the needs of low cost, small form factor, and high performance. For the applications of these packages, IC silicon wafers have to be ground to be relatively thin through the wafer-thinning processes (such as grinding, polishing, and plasma etching). The strength of dies has to be determined for the design requirement and reliability assurance of the packages. From the published data, there still exist some issues, including a large scatter existed in die strength data and difficulties in differentiating the causes of the low strength between from the wafer grinding and from wafer sawing by either the three-point bending or four-point bending test. The purposes of this study are to develop new, reliable, and simple test methods for determination of die strength, in order to improve the data scatter, and to provide a solution for differentiating the factors that affect the variability of die strength for finding out the causes of the weakness of the die strength. In this study, two new test methods, point-loaded circular plate with simple supports test (PLT-I) and point-loaded plate on elastic foundation test (PLT-II), are proposed and then evaluated by testing two groups of silicon dies with different surface conditions. The surface conditions (roughness) of the specimens are determined by atomic force microscopy and correlated to failure strength. The failure forces from both tests have to be modified by using maximum stress obtained from theories or finite element analyses to obtain the failure strength. The test results are compared to each other and further with a widely used four-point bending test. The results suggest that, unlike the four-point bending test suffering the chipping effect, both methods provide very consistent data with a small scatter for each group of specimens and can be used for identifying the effect of surface grinding (roughness) on the die strength. It is also shown that the die strength is highly dependent on the surface roughness. Accordingly, these two methods can provide not only a (biaxial) stress field similar to temperature-loaded die in the packages, but also simple, feasible, reliable, and chipping-free tests for silicon dies of dummy or real IC chips, without strict geometrical limitation, such as beam-type geometry for the three-point or four-point bending test.