Multi-view neural 3D reconstruction of micro- and nanostructures with atomic force microscopy

Area of Science:

• Nanotechnology
• Surface Science
• Microscopy

Background:

• Atomic Force Microscopy (AFM) is crucial for micro- and nanoscale topographic imaging.
• Conventional AFM faces challenges in precisely reconstructing complex 3D structures due to artifacts and incomplete data.
• Existing 3D-AFM methods often require specialized probes or system modifications.

Purpose of the Study:

• To develop an accurate and cost-effective framework for reconstructing complex 3D micro- and nanostructures using AFM.
• To overcome limitations of conventional AFM, including tip-sample convolution artifacts and incomplete topography capture.
• To introduce a novel approach that does not necessitate specialized AFM probes or hardware modifications.

Main Methods:

• Proposed a multi-view neural-network-based framework integrated with AFM (MVN-AFM).
• Employed an iterative method for multi-view data alignment and simultaneous artifact elimination.
• Applied neural implicit surface reconstruction techniques to nanotechnology applications.

Main Results:

• MVN-AFM accurately reconstructs surface models of intricate micro- and nanostructures.
• The framework effectively eliminates artifacts present in raw AFM images.
• Successfully reconstructed diverse nanostructures, including two-photon lithography microstructures, PMMA nanospheres, and ZIF-67 nanocrystals.

Conclusions:

• MVN-AFM offers a cost-effective and accurate solution for 3D micro- and nanoscale analysis.
• The developed framework enhances the capabilities of AFM for complex structure imaging.
• This approach represents a significant advancement in nanoscale metrology and surface characterization.