A general machine learning (ML) framework of surface wetting is proposed by considering a broad range of factors, including solid surface topography, solid surface chemistry, liquid properties, and environmental conditions. In particular, an XGBoost-based ML model is demonstrated for learning the surface wetting behaviors processed by a laser-based surface functionalization process, namely nanosecond laser-based high-throughput surface nanostructuring (nHSN). This is the first known attempt to apply machine learning to surface wetting by considering both surface topography and surface chemistry properties. Novel microscale and nanoscale topography parameters viz., roughness, fractal, entropy, feature periodicity are defined with suitable computer algorithms to comprehensively describe the surface topography. A novel set of surface chemistry parameters such as polarity, volume, and amount of functional groups are also used as the machine learning model input. Upon analyzing the importance of each parameter for the nHSN process, surface chemistry shows the greatest importance in determination of surface wettability, while surface morphology also plays a part in influencing the wettability. • A machine learning model is firstly proposed for surface wettability considering both topography and chemistry. • Fractal dimension, 2D-entropy, and periodicity are used to describe surface morphology. • Polarity and functional groups are used to describe surface chemistry. • The results showed that the surface chemistry is the dominant feature while surface topography plays a supportive role.
Read full abstract