Graphene oxides (GO) are nanosheets of graphene with oxygen-bearing defects. The GO sheet is highly heterogeneous with the sp3 carbons, connected to oxygen-bearing functional groups, interspersed between the usual sp2 carbons of graphene, resulting in competition between aromatic regions and highly hydrophilic domains (from the oxygen-containing functional groups, typically -OH and epoxy groups) at the GO-water interface. GO materials are increasingly gaining traction in separation science. The interfacial region between the GO material and water plays a key role in these applications. To probe the interfacial region with a high level of accuracy, ab-initio molecular dynamics (AIMD) simulations are necessary. However, due to the high computational cost, these AIMD simulations are constrained in both length and time scales. Machine learning-based force-fields can help to solve this accuracy-cost dilemma. GO-water interfaces at different oxidation levels of GO (GO_2/1, GO_4/1, GO_8/1, Graphene) are studied using deep neural network-based force fields. The force-field was developed by training a deep neural network model on ab-initio molecular dynamics simulation data from GO-water interfaces at two different oxidation levels, namely the fully oxidized GO_2/1 with a C/O ratio of 2:1 and a partially reduced form GO_4/1, with a ratio of 4:1. The force-field was then validated on more reduced sheets including the limiting case of graphene-water. The interfacial liquid structure, orientation, solvating properties, wettability, and reactivity at the GO/water interfaces, as a function of oxidation level, will be presented. Furthermore, the signature features of the sum frequency generation spectra for the different cases will be presented and analyzed based on the interfacial structure and reactivity.

Rolf David

Appearances