Dual-indicators machine learning assisted processing high-quality laser-induced fluorine-doped graphene and its application on droplet velocity monitoring sensor

Abstract

Laser-induced graphene (LIG) is widely used in electron devices owing to its characteristics of fine patterning and high precision. Doping is a commonly used approach to improve the quality of LIG. Thanks to the excellent superhydrophobicity and high electrical conductivity of laser-induced fluorine-doped graphene (F-LIG), it shows promising potential in developing electrodes for electron devices. However, it remains a major challenge to rapidly search for the optimal laser processing parameter space that can process high-quality F-LIG. In this work, a dual-indicators predictive machine learning (ML) method for optimizing the F-LIG process was proposed, utilizing a Gaussian process regression (GPR) algorithm. The mean square error (MSE) between the actual and predicted values of the data used for the model test was 0.00814, while the mean absolute percentage error (MAPE) was 10.33 %, demonstrating a certain degree of generalization without overfitting. Importantly, the F-LIG processed using parameters predicted by the ML model achieved excellent superhydrophobicity and conductivity simultaneously. Based on this high-quality F-LIG, a self-powered droplet velocity monitoring sensor was developed. This sensor can accurately work after 20,000 cycles, showing excellent stability. The proposed dual-indicator ML model assisted the optimization of the F-LIG process, which will provide a feasible and economical way to process high-quality graphene and related electron devices.