Influence of the Thickness of Silicon Dies on Strength

Abstract

The market share of thin semiconductors has continuously increased in microelectronical, micromechanical as well as in the solar industries in the recent years, e.g. due to required flexibility for RFIDs or cost reduction of solar cells. However thin wafers are difficult to handle, because of the increasing flexibility and increasing sensitivity to mechanical, thermal and intrinsic loads in manufacturing and use. Therefore the mechanical properties, especially strength, have to be investigated in order to optimize manufacturing steps with regard to the reliability. In the semiconductor industry one can find a lot of reports about the decreasing strength of thin silicon devices. The small thickness seems to be responsible for early fracture in manufacturing. In this work, the strength of thin silicon is investigated. For the investigation (3times3)mm2-dies with a thickness between 200mum and 48mum made from (100) single crystalline silicon were investigated using the ball on ring test. All specimens were thinned back by grinding and wet-chemical spin-etching for stress relief. The front side was not treated by an additional process. In ball on ring tests the maximum principle stress occurs at the surface at the center of the specimen. Since the stress at the edge of the sample is significantly smaller than the stress at the sample center the fracture starts in the center of the sample. Thus the influence of the back thinning technology can be characterized and the dicing process does not influence the test results. For statistical evaluation 40 specimen of each thickness were tested. The front side was also tested as reference. Weibull theory, based on the weakest link model, was chosen for statistical evaluation. Due to the small thicknesses of the samples the force-displacement curves show a nonlinear relationship. Hence the finite element method in consideration of large deflection (geometric nonlinearity) was applied to calculate the fracture stress from the fracture force of each specimen. Additional the contact behavior (structural nonlinearity) between ball and specimen was modeled to consider the changing boundary conditions in large deflection. The influence of the chip thickness on the characteristic fracture stress is shown. It can be seen, that the strength is increasing with decreasing sample thickness for both front and back side. The fracture stress increases very strongly in the range of 50...100mum. It has to be kept in mind that all samples were treated with same process steps. Thus it can be assumed that all samples show the same flaw size and flaw distribution. Hence it can be concluded, that the strength of identical manufactured samples depends on the sample thickness for small thicknesses. In the case of very small sample thickness (less than 20mum), some specimen showed buckling in ball on ring test, caused by large deflection. In order to derive reliable stress values in numerical calculation this process of instability has to be investigated. With the results guidelines can be given for reliable ball on ring testing of thin silicon samples. The influence of the thickness of silicon dies on the strength properties as well as challenges in testing are the focus of further investigations