Preparation and characterization of Ti3C2 Mxene through back propagation neural network-genetic algorithm combined with surface method response

Abstract

Mxene's distinctive two-dimensional (2D) layered structure and special surface chemistry make it highly versatile for adsorption, sensors, elecatalysis, and energy storage. However, it has always been a challenge to obtain Mxene with excellent performance. In this study, the process of Ti3C2 Mxene preparation by etching Ti3AlC2 with hydrofluoric acid was optimized using the back-propagation neural network-genetic algorithm (BPNN-GA) and response surface methodology (RSM). Compared to RSM, BPNN-GA predicted better accuracy for preparation conditions with higher values for R2 (0.9830), MSE (2.6188), RMSE (1.6183), MAE (0.8504), and MAPE (0.0089 %). The BPNN-GA optimized etching values for HF concentration, time, and temperature were 8.81 %, 7.01 h, and 45.07 ℃, respectively. Under these conditions, the Ti3C2 Mxene prepared achieved 99.39 % removal efficiency of methylene blue (MB). The preparation was validated by XRD, SEM, TEM, XPS, Zeta, and FTIR analyses, indicating Ti3C2 Mxene to be a two-dimensional crystal structure with a multilayer arrangement and surface termination groups. Zeta measurements indicated that its surface carried negative charges across a wide pH range and its suitability for cationic dye adsorption. Using BPNN-GA provides valuable guidance and a novel strategy for material preparation.