Machine-Learning Accelerated Computational Spectroscopy Reveals Vibrational Signature of the Oxidation Level of Graphene in Contact with Water

Precise characterization of the graphene-water interface has been hindered by the experimental inconsistencies and limited molecular-level access to interfacial structures. In this work, we present a novel integrated computational approach that combines machine-learning-driven molecular dynamics simulations with first-principles vibrational spectroscopy calculations to reveal how graphene oxidation alters the interfacial water structures. Our simulations demonstrate that pristine graphene leaves the hydrogen-bond network of interfacial water largely unperturbed, whereas graphene oxide (GO) with surface hydroxyls induces a pronounced Δν̃ ≈ 100 cm^-1 redshift of the free OH vibrational band and a dramatic reduction in its amplitude. These spectral shifts in the computed surface-specific sum-frequency generation spectrum serve as sensitive molecular markers of the GO oxidation level, reconciling previously conflicting experimental observations. By providing a quantitative spectroscopic fingerprint of GO oxidation, our findings have broad implications for catalysis and electrochemistry, where the structuring of interfacial water is critical to the performance.