Water Molecule Diffusion in Graphene Oxide: Exploiting Machine Learning Algorithms for Advantages and Insights

Shuo Huang,Li Zeng,Zhiwei Zhou
DOI: https://doi.org/10.1016/j.commatsci.2023.112769
IF: 3.572
2024-01-01
Computational Materials Science
Abstract:Graphene oxide (GO) membranes have attracted much attention due to their wide applications in areas such as desalination and water filtration. In this work, we have compared four machine learning (ML) algorithms (Bagging, XGBoost, AdaBoost, and RF) with molecular dynamics simulation (MD) results as the database to predict the water diffusion affected by crucial GO structure features such as the oxidation degree and interlayer spacing. The machine learning results suggest that the model based XGBoost algorithm has the best performance among the four models. Moreover, RF-based machine learning algorithms assessed the importance of two structural features of GO on water diffusion. The results indicate that the layer spacing has a significantly greater effect on the water diffusion than the oxidation degree within a 30% oxidation degree and a 2-nm layer spacing. The hydrogen bonding between the water and oxygen groups on the GO surface was chosen to investigate the mechanism behind the phenomenon. This work can help us predict the water diffusion and understand the effects of the structure properties of GO membranes on water treatment, which can aid in the design of graphene oxide nanofiltration.
What problem does this paper attempt to address?