Machine learning-aided 3D-AFM for identification of spatial heterogeneity in interfacial solvation structures.

Area of Science:

• Surface Science
• Materials Science
• Nanotechnology

Background:

• Interfacial solvation structures are critical for applications like lubrication and energy storage.
• Three-dimensional atomic force microscopy (3D-AFM) offers angstrom-precision mapping of liquid-solid interfaces.
• Conventional analysis of 3D-AFM data struggles with massive datasets, hindering the study of spatial heterogeneities.

Purpose of the Study:

• To develop a general method for unbiased interpretation of 3D-AFM interfacial force datasets.
• To leverage unsupervised machine learning for analyzing large-scale interfacial data.
• To uncover and characterize spatial heterogeneities at liquid-solid interfaces.

Main Methods:

• Integration of 3D atomic force microscopy (3D-AFM) with unsupervised machine learning algorithms.
• High-throughput analysis of massive interfacial force datasets.
• Molecular dynamics simulations to interpret observed interfacial phenomena.

Main Results:

• A novel method for unbiased interpretation of 3D-AFM data was successfully developed.
• Spatial heterogeneity at the graphite-water interface was identified and characterized.
• The interface was partitioned into two distinct regimes with differing mechanical responses, attributed to hydrocarbon contaminants.

Conclusions:

• Unsupervised machine learning effectively handles large 3D-AFM datasets, enabling objective analysis of interfacial structures.
• Contaminant-mediated modulation of interfacial water structures significantly impacts interface properties.
• This work provides a general framework for detecting chemical non-uniformities at solid-liquid interfaces.