The understanding of the origin of the intrinsic stress during the formation of polycrystalline thin films and coatings is a topic with deep technological implications, because residual intrinsic stress endures after the film is incorporated into a device. The accumulation of this stress under operating conditions causes the thermomechanical fatigue that is estimated to be responsible for 90% of mechanical failures in current devices.
The mechanisms responsible for the intrinsic stress of compression during postcoalescence (the growth stage where the polycrystalline film becomes continuous) have generated three-decade-long intense discussion without a consensus so far. The difficulty of reaching a consensus lies in the fact that current models of stress generation, most of them atomistic, attempt to explain the experimental evidence in terms of macroscopic average behaviors, while the microscopic stress distribution is unknown.
In this work, we directly measure the local distribution of residual intrinsic stress in polycrystalline films on nanometer scales, using a pioneering method based on atomic force microscopy [C. Polop et al., Nanoscale 9, 13938 (2017)]. Our results demonstrate that, at odds with expectations, compression is not generated inside grain boundaries but at the edges of gaps where the boundaries intercept the surface. We propose a model of competition between surface diffusions (namely, Mullins-type diffusion driven by surface gradients of the dangling bond density vs. Srolovitz-type diffusion induced by gradients of the surface strain) to explain the origin of the postcoalescence compression as well as its evolution with the film thickness and dependency on the deposition flux. This model addresses comprehensively all of the major experimental findings that have been reported up to date concerning the kinetics of the intrinsic stress in polycrystalline films. [Full article]