A generative diffusion-based AI-model to scrutinise the microstructural degradation of isothermally aged Cu-SAC305 interfaces

Abstract Generative artificial intelligence (genAI) has disrupted multiple disciplines, as it has the potential to drastically change the way content is created. In this study, microstructural grey-value images of thermally aged interfacial Cu–SAC305 microstructures are virtually generated for seen, unseen and even unknown sample conditions utilising a denoised diffusion probabilistic model (DDPM). The DDPM is conditioned based on the different ageing time and impurity levels. The multi-scale structural similarity index is calculated to assess the quality of generated images. Kirkendall pore areas and IMC-thicknesses are extracted from labelled real and generated images as physical descriptors, to quantitatively characterise the microstructure. Both follow the expected physical trends of parabolic growth, confirming the underlying diffusion process and validating the pertinency of the model. Hence, this study demonstrates the feasibility of genAI to significantly enhance systematic imaging investigations of microstructural degradation with concurrently reducing experimental effort.