Machine Learning in optimizing Carbon Nanotubes and Graphene Production

Abstract

Machine learning plays a key role in this work, designing new technologies for producing carbon nanotubes (CNTs) and graphene by electrolysis in molten salts. The aim is to achieve non-expensive, high-quality materials, making them economically viable for various applications. For the production of multi-walled carbon nanotubes (MWCNTs), experiments employ both non-stationary and stationary current regimes, while for graphene production, constant and reversing cell voltage as well as constant and reversing overpotential methods are considered. The electrolysis process offers ecological and economical advantages with precise control over parameters such as applied voltage, current density, temperature, electrolyte type, and graphite material. To determine the relationship between these parameters and material quality, explainable tree-based Machine Learning (ML) models are employed, trained using labeled data from domain experts. The extracted rules from the ML model guide optimal production, resulting in high-yield materials that are up to ten times more cost-effective than existing technologies. This contributes to the advance of cost-efficient and high-quality carbon nanomaterials for a wide range of applications.