Composition-Controlled Hydrogen-Bond Restructuring and Interfacial Transport in Glycerol–Ethanol Mixtures for Orange Essential Oil Separation: Experiments and Multiscale Modeling

Hydrogen-bonded glycerol–ethanol (Gly–EtOH) mixtures are promising biodiesel-derived solvents for orange essential oil (OEO) separation. However, how solvent composition governs equilibrium and transport properties and selectivity remains unclear. We combined rotational rheometry, ReaxFF molecular dynamics (ReaxFF-MD), density functional theory (DFT), COSMO-RS, Redlich–Kwong equation of state (RK EoS), and machine learning to establish structure–property–separation relationships. ReaxFF-MD and RK EoS reproduced the composition-dependent density behavior with AARDs of 4.6% and 2.0%, respectively. Rheometry validated the ReaxFF-MD-predicted decrease in viscosity with increasing ethanol content. ReaxFF-MD and DFT reveal that ethanol reorganizes the glycerol-dominated hydrogen-bond network by altering O···O coordination and O···H correlations, while COSMO-RS reveals a composition-dependent shift toward weakly polarized surface segments. Polynomial ridge regression and Gaussian process regression were used to interpolate density and viscosity, respectively. Source-separated and grouped cross-validation showed that composition interpolation was more reliable than prediction at a completely unseen temperature. Interfacial MD simulations further indicate that solute–solvent affinity, interfacial organization, and diffusivity contrast jointly govern linalool enrichment, with the highest simulated selectivity at x1 = 0.5. These findings establish a composition-controlled mechanism connecting hydrogen-bond restructuring, bulk transport, and interfacial separation, enabling rational design of sustainable biomass-derived solvents.