Predicting water dynamics on nanopatterned surfaces using convolutional neural network algorithm

Nanopatterns are widespread on surfaces and play a crucial role in water dynamics, which is particularly significant in the fields of surface science, electronic devices, and advanced materials. However, the relationship between nanopatterned surfaces and water diffusion is complex and requires further research. In this paper, we use a convolutional neural network (CNN) algorithm in conjunction with molecular dynamics (MD) simulation results as a database to predict the water diffusion behavior influenced by the nanopatterned distribution of oxygen-containing groups on graphene oxide (GO) surfaces. The results show that the CNN algorithm performs effectively in predicting water diffusion on regularly nanopatterned GO surfaces, achieving a Pearson correlation coefficient (r) of 0.984 and a coefficient of determination (R2) of 0.967. Furthermore, this trained CNN model was extended to predict the diffusion of water molecules on unseen irregularly nanopatterned GO surfaces, yielding an r of 0.956 and an R2 of 0.857, indicating promising but preliminary generalization capability. This work presents a novel methodology for predicting the water dynamics on nanopatterned surfaces, aiding in the design of micro- and nano-scale flow surfaces.