Deep Learning strategies recently emerged as powerful tools for the characterization of heterogeneous materials. In this work, we discuss an approach for the characterization of the mechanical response of polysilicon films that typically constitute the movable structures of microelectro-mechanical systems (MEMS). A dataset of microstructures is digitally generated and a neural network is trained to provide the appropriate scattering in the values of the overall stiffness (in terms of the Young's modulus) of the grain aggregate. Since results are framed within a stochastic procedure, the aim of the learning strategy is not to accurately reproduce the microstructure-informed response of the polysilicon film, but instead to provide a fast tool to be used at the device level for Monte Carlo analysis of the relevant performance indices. Accuracy of the proposed approach is assessed for very small samples of the polycrystalline aggregate to check if size effects are correctly captured.
Background: Due to their size, Micro Electromechanical Systems (MEMS) display performance indices affected by uncertainties linked to the mechanical properties and to the geometry of the films constituting their movable parts. Objective: In this perspective, a recently proposed multiscale and hybrid solution for uncertainty quantification is discussed. Methods: The proposed method is based on the (deep) learning of the morphology-affected elasticity of the polycrystalline films, and of the microfabrication-induced defective geometry of the devices. The results at the material and at the device levels are linked through a reduced-order representation of the response of the entire device to the external stimuli, foressen to finally feed a Monte Carlo uncertainty quantification engine. Results: Preliminary results relevant to a single-axis resonant Lorentz force micro-magnetometer have shown a noteworthy capability of the proposed multiscale deep learning method to account for the mentioned uncertainty sources at the microscale. Conclusion: A promising two-scale deep learning approach has been proposed for polysilicon MEMS sensors to account for both materials- and geometry-governed uncertainties, and to properly describe the scale-dependent response of MEMS devices.
Smart materials and structures, especially those bio-inspired, are often characterized by a hierarchy of length-and time-scales. Smart Micro Electro-Mechanical Systems (MEMS) are also characterized by different physical phenomena affecting their properties at different scales. Data-driven formulations can then be helpful to deal with the complexity of the multi-physics governing their response to the external stimuli, and optimize their performances. As an example, Lorentz force micro-magnetometers working principle rests on the interaction of a magnetic field with a current flowing inside a semiconducting, micro-structured medium. If an alternating current with a properly set frequency is let to flow longitudinally in a slender beam, the system is driven into resonance and the sensitivity to the magnetic field may result largely enhanced. In our former activity, a reduced-order physical model of the movable structure of a single-axis Lorentz force MEMS magnetometer was developed, to feed a multi-objective topology optimization procedure. That model-based approach did not account for stochastic effects, which lead to the scattering in the experimental data at the micrometric length-scale. The formulation is here improved to allow for stochastic effects through a two-scale deep learning model designed as follows: at the material scale, a neural network is adopted to learn the scattering in the mechanical properties of polysilicon induced by its polycrystalline morphology; at the device scale, a further neural network is adopted to learn the most important geometric features of the movable parts that affect the overall performance of the magnetometer. Some preliminary results are discussed, and an extension to allow for size effects is finally foreseen.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.