Simulation-based study demonstrates high classification accuracy for tip shapes in dynamic atomic force microscopy, implying advances in imaging techniques.
Key Points
This research aims to develop a classification framework to identify probe tip shapes in dynamic atomic force microscopy using simulated data.
Established a dimensionless dynamic model of the microcantilever
Extracted multidimensional dynamic feature parameters from interaction force models
Constructed an XGBoost-based classifier and used SHAP for interpretability
Non-contact classification accuracy reached 96.7%, with 100% for conical tips
Under tapping-mode, classification rates were 100% for conical, 85.5% for spherical, and 98.3% for flat tips
Demonstrated feasibility of identifying tip shapes based on dynamic responses via simulated data