Data-driven optimization for controllable multi-scale aperture fabrication of nanopipettes

This study addresses the critical challenge of controllable nanopipette fabrication by proposing a multi-parameter collaborative optimization framework that combines an artificial neural network (ANN) with a physical model. The ANN was utilized to establish a nonlinear mapping between the five strongly coupled fabrication parameters (Heat, Filament, Velocity, Delay, Pull) and nanopipette aperture size. Three complementary feature weight analysis methods-Random Forest (RF), SHAP values, and the Garson algorithm-were employed to rank the importance of these parameters, which consistently identified Heat as the most influential parameter, followed in that order by Pull, Delay, Filament, and Velocity. Concurrently, a physical model characterizing the temporal evolution of nanopipette aperture during the pulling process was derived from mechanical theory. Under the guidance of the theoretical model and parameter importance ranking, successful controllable fabrication of nanopipettes with target apertures spanning 50 nm to 1000 nm in 100 nm increments was achieved. This work thus transforms nanopipette fabrication from an empirical trial-and-error approach into a predictable, model-driven paradigm.