Two main uncertainty sources can affect the response of polysilicon MEMS during standard on-chip measurements: the overetch induced by the deep reactive-ion etching process; the mechanical properties of the aggregate of silicon grains. The former one can be reduced by finer fabrication techniques, not adopted indeed in mass production processes, while the latter one is related to the length-scale of the devices.
Due to the increasing miniaturization, the width of some MEMS components may become comparable to that of a silicon grain and the relevant effective mechanical properties can vary significantly from one device to another. In this work, through on-chip tests we investigate the response of polysilicon films using standard electrostatic actuation/sensing. The results of such experimental campaign are then compared to an analytical reduced-order model of the structure, and to coupled electro-mechanical simulations accounting for the stochastic morphology of the silicon film. These two models are adopted to bilaterally bound the experimental data up to pull-in, and to assess the scattering induced by the random orientation of the crystal lattice of each grain in digital Voronoi tessellations of the slender parts of the devices.
I think you are raising an interesting point here - from my mechanical engineering practice, I am used to assume that polycrystalline materials may be treated as isotropic in most cases, since grain orientations average out, but in your case, I understand that both the size of structures and the dependency of etch rates on lattice orientation make it important to consider the property variations that occur as a result. Nice approach, highlighting the issue.
Kind regards,
Dirk Lehmhus
Besides the numerical modeling of grains explicitly, as bending deformation (unlike the tension) produces strain gradients in the beam, the higher order elasticity models can be useful to describe the mechanical response analytically.
Best regards,
Ramin Mirzazadeh