Total: 1
Micro- and nanostructures fabricated by focused ion beam (FIB) milling invariably deviate from input designs due to systematic and stochastic effects such as ablation-rate fluctuations, and redeposition. As such, nominally identical iterations of a given process yield a distribution of non-identical fabrication outcomes, leading to variations in the physical (optical, mechanical) properties of fabricated structures. Existing models of the milling process typically predict a single deterministic fabrication result, neglecting this variability. Here we present a conditional diffusion model that learns the statistical distribution of process outcomes conditioned on input design, enabling it to predict a realistic distribution of possible output sample structures. These predictions accurately reproduce edge morphology, linewidth variations, and defect probabilities observed in fabricated samples, enabling statistical analysis of milling performance and the identification of dominant structural and process parameter-dependent deformation modes. The work establishes a probabilistic modelling framework for fabrication variability analysis and fabrication-aware design optimization.