Warpage of micro-electronics caused by the curing process and thermal cycling is of major importance in electronic packaging. Industry is availed by good methods to be able to predict warpage accurately. The main difficulty for prediction of warpage is caused by the complicated material behavior of molding compound. It turns out that the mechancial behavior of molding compound is dependent on time, temperature and degree of conversion. Since molding compound is available in large variabilities, for each type the model parameters should be established experimentally. In our group an efficient method is developed for experimentally determining the model parameters [1, 2, 3]. In this approach, different experimental results have to be combined in order to achieve the final material model. Since this material model is not standard, user-subroutines are used for implementation in finite element software (ABAQUS). In this paper we present validation experiments which are done in order to verify the material model. A TDM (Topography and Deformation Device) is used to measure the curvature of a mold-map at different temperatures. Good agreement between experiment and simulations results is achieved.@en

Thermally conductive adhesives (TCA) and electrically conductive adhesives (ECA) are one of the major concerns of the contemporary micro-electronics. They are especially important in application where, e.g. effective heat dissipation is the key factor for reliability issues. Currently there is a lot of ongoing research in order to improve the basic thermal property of adhesives, which is mainly heat conductance. According to the literature data the heat conductance can vary from 0.1 up to 60 W/m K. It depends not only on the filler material, its content and configuration but also on thermo-mechanical properties of matrix. On the other hand numerical simulation becomes nowadays an inevitable tool for rapid non-destructive and low-cost experiments. The basic problem of numerical experiments is accuracy. Nevertheless the accuracy can be improved by combining the numerical and traditional experiments. This can be achieved by means of partial validation of numerical results by traditional experiments or by precise and appropriate material properties measurement. In fact, the above approach was applied in current work in order to simulate the influence of curing temperature and time on the thermal conductance of polymers. Thermally conductive adhesives belong to polymer based materials. In order to apply numerical simulation it is required to have an appropriate description of the thermal and mechanical behavior of polymers. Most often polymers are described by cure dependent or independent linear viscoelastic model [3] and [5]. Having this model, which parameters in fact can be measured experimentally, it is possible to simulate the stress and strain field caused by polymer curing and shrinkage phenomena and finally assess the thermal conductance accordingly. Current paper focuses on a problem of numerical simulation of TCA in order to recognize the trend dependence of thermal conductivity due to viscoelastic model of polymers and filler particles contact area.@en

Commercial FEM software codes do not directly support the cure dependency of polymers behavior. In electronics packaging, for example, cure shrinkage of the molding compound is consequently treated as a surplus to thermal contraction and its changes in visco-elastic behavior during curing are usually ignored. This gives rise to inaccurate results and may cause difficulties even in recognizing factors important for further optimization. This work introduces a cure dependent visco-elastic material model being implemented into ANSYS by USERMAT, the user material subroutine. The results obtained when simulating the warpage of bi-material strips during and after curing with the full scale material model were compared to those obtained with simplified material models. Varying the geometric configuration different changes in the deformation results have been found caused by cure dependency anywhere between -10%, 0%, and +20%. Hence, correct results can only be achieved in general case with models accounting for cure dependency directly.@en