SUMMARY

The thesis is based on the study of one of the most frequent failure mechanisms in semiconductor packages, the delamination of interface which is the separation of two bonded materials, in order to improve their adhesion and a fortiori the reliability of microelectronic devices. It focuses on the metal/polymer interfaces because they cover 95% of all existing interfaces. Since several years, research activities have proved that the more roughened the surface of the interface, the better the adhesion between these two materials. However, the roughness exhibits extremely complex shapes which make it difficult to treat analytically or numerically. In order to investigate quantitatively the effect of roughness variation on adhesion properties, studies have been carried out involving analytical fracture mechanics, then numerically with Finite Element Analysis in a deterministic way by assuming an ideal profile which is repeated periodically. However, on the one hand, with the development of statistical and stochastic methods representing roughness as random process and field, and on the other hand, with the emergence of probabilistic fracture mechanics, the present work adds a stochastic framework to the previous studies. One of the Stochastic Finite Element Methods, the Perturbation method is interesting, because instead of having many Finite Element models for each type of interface shape, it leads equivalently to only one model. Indeed, it can carry out at once what traditional Finite Element Analysis does with numerous simulations which requires changing geometric parameters each time. This method is implemented as a module in a Finite Element package. Applying the Perturbation method to the roughness is too vague at this stage, because no optimal representation of the latter has been done yet. Therefore, the 3 point bending test on a beam problem has been developed analytically and numerically in order to validate the Perturbation method.