PeakForce AFM Analysis Enhanced with Model Reduction Techniques.

Area of Science:

• Materials Science
• Nanotechnology
• Data Science

Background:

• PeakForce quantitative nanomechanical Atomic Force Microscopy (PF-QNM) generates high-dimensional datasets for mechanical property analysis.
• Analyzing heterogeneous materials with complex topography presents segmentation challenges.
• Existing methods often require prior mechanical models and can be subjective.

Purpose of the Study:

• To develop a novel data processing pipeline for PF-QNM data.
• To reduce the dimensionality of PF-QNM datasets using proper orthogonal decomposition (POD) and machine learning.
• To enable objective and efficient extraction of underlying mechanical parameters.

Main Methods:

• Application of proper orthogonal decomposition (POD) for dimensionality reduction.
• Utilizing machine learning techniques on the reduced-dimensionality data.
• Investigating heterogeneous samples: polystyrene with nano-pods and PDMS with particles.

Main Results:

• Successful compression of high-dimensional PF-QNM data into a lower-dimensional representation.
• Extraction of key 'state variables' governing mechanical responses.
• Demonstrated interpretation of material phases, interfaces, and topography from mechanical data.

Conclusions:

• The proposed method significantly reduces user dependency and subjectivity in data analysis.
• It offers a compact and straightforward interpretation of complex force-indentation data.
• The approach is computationally efficient and model-agnostic.