Semiconductor devices can be found everywhere in our daily lives, for example in self-driving cars, bank cards and personal devices such as smart phones or notebooks. Once part of these personal devices, one does not want them to show failure. Although reliability of such devices is taken into account in the design, the fabrication process might lead to the emergence of small damages in the product. Since these damages or their propagation might cause failure of the device, a series of visual inspections and functional tests is part of the fabrication process. The obtained resolution by optical microscopy, currently the state of the art inspection method, is not sufficient for the detection of small damages such as microcracks or damages located inside a sample. To keep improving the reliability of semiconductor devices, these damages need to be detected in another high speed, low cost way.

As semiconductor devices shrink in size, their natural vibration frequencies increase and approach the
MHz-range. Vibration based damage detection methods might therefore offer an alternative high speed in-situ inspection method. The main goal of this thesis is to identify and experimentally verify vibration characteristics that indicate the presence of damage, with a focus on microcracks, in semiconductor devices. While linear vibration based damage methods have proven to be insensitive to small damages such as microcracks, non-linear vibration based damage detection methods show much higher sensitivities to this type of damage. The non-linear elastic wave spectroscopy (NEWS) of several damaged materials have shown two characteristic phenomena: amplitude dependent natural frequency shifts and the generation of higher harmonics. Both are explained by a phenomenological non-linear hysteric elastic model. While the applicability of NEWS is proven for several materials, its performance for silicon, in particular at microscale, is still unknown.