Data-driven optimisation of eco-friendly graphene production via electrolysis in molten salts using decision tree model

Abstract

This study presents a comprehensive investigation into the sustainable production of high-purity graphene via molten-salt electrolysis combined with machine-learning optimisation. Graphite was electrochemically exfoliated in molten sodium hydroxide under non-stationary current regimes, reversing current, pulsating current, and pulsating overpotential, to achieve controlled intercalation and exfoliation. The effects of graphite type, cell voltage, electrolyte temperature, cathodic overpotential, and polarity-change time on graphene morphology and yield were evaluated across 30 experimental runs. The resulting graphene showed high crystallinity, few-layer stacking, and minimal defect density, verified by Raman spectroscopy, X-ray diffraction, scanning/transmission electron microscopy, and thermogravimetric/differential thermal analysis, with purity exceeding 99.4% as confirmed by inductively coupled plasma–optical emission spectroscopy. A decision tree (DT) model was developed to classify graphene quality into three levels and to identify dominant synthesis parameters, revealing graphite type, cell voltage, and electrolyte temperature as key factors controlling product quality. By enabling interpretable, rule-based prediction of synthesis outcomes, the DT model reduced empirical trial-and-error and supported targeted optimisation. Integrating experimental electrochemistry with data-driven modelling, this work establishes a scalable, eco-efficient route for graphene production and highlights molten-salt electrolysis as a metal-free, sustainable platform for large-scale manufacturing, with DT learning providing a powerful tool for process control and materials design.