Ultra-thin polymer films are increasingly utilized in next-generation freeform electronics due to their inherent conformability. However, an ultra-thin polymer film under stretching shows nonlinear mechanical behavior accompanied by stochastic nanoscale damage. Thus, diagnosing the stretched, sub-100 nm thick polymer films is essential to ensure the reliability of this new type of damage-tolerant design. However, the conventional approaches were limited by the intrinsic complexity of the testing platform and the low dimensionality of the correlation strategy. Here, a knowledge-guided framework of strain-damage image correlation (SDIC) that integrates the film-on-water (FOW) tensile testing and the artificial intelligence (AI) regression model is presented. High-fidelity strain and damage image data were prepared by freely stretching ultra-thin polystyrene (PS) films on water surface with in situ observation, and consequently correlated in a high-dimensional latent space. Importantly, the acquired raw in situ image data were processed in the guidance of domain knowledge based on polymer science and solid mechanics. After confirming the assumption for dimension reduction with experiments and preventing the spurious correlation with simulations, an improved SDIC was established by introducing a multiple cropping strategy for statistically evolving damage. This framework is expected to further elucidate the complex behaviors of various ultra-thin polymer films and ensure the mechanical reliability of future stretchable electronics. • High-fidelity data of ultra-thin polymer films were acquired on FOW platform. • Knowledge-guided data processing was performed. • High-dimensional strain-damage image correlation was accomplished with AI.
Song et al. (Sun,) studied this question.