Predictive Modeling of Thin Film Yield Stress Using Machine Learning: A Simulation-Based Approach

In this study, an experimental and simulation-based approach is used to evaluate the yield stress of integrated circuit (IC) industry-relevant thin film materials. An artificial neural network (ANN) is trained to predict the yield stress of thin films based on the load-displacement curves obtained from the finite element simulations of thin film nanoindentation (NI). The NI models include a wide range of film thickness values (100-500 nm) as well as elastic and plastic properties for the thin film material. The validity of the yield stress model is assessed through a series of thin film NI experiments and test simulations. The results show that the machine learning approach enables a rapid assessment of the yield stress for thin film materials, and the ANN model demonstrates accurate yield stress predictions with an error rate of 3.59%.