Characterization of ibuprofen removal by calcined spherical hydrochar through adsorption experiments, molecular modeling, and artificial neural network predictions.

ABSTRACT

Ibuprofen (IPF) is one of the most prescribed nonsteroidal anti-inflammatory drugs in recent times, but it is not readily removed in conventional wastewater treatments. Here, we investigate the adsorption characteristics of IPF onto calcined spherical hydrochar (CSH), which was synthesized through hydrothermal carbonization of sucrose followed by calcination. The adsorption experiments show that the equilibration time for IPF was 360 min, and a pseudo-second-order model was best fitted to the kinetic data. The isotherm data were best described by the Liu model with a theoretical maximum adsorption capacity of 95.6 mg/g. The thermodynamic data indicate the endothermic nature of the adsorption at 10-40 °C. The CSH was favorably regenerated and reused using methanol. In pH experiments, the IPF adsorption capacity declined gradually as pH rose from 2 to 8, dropped rapidly at pH 10, and became negligible at pH 12. The IPF adsorption to the CSH could occur through various adsorption mechanisms. Hydrogen-bond formation, π-π interactions, n-π* interactions, and electrostatic repulsion were explored and visualized with molecular modeling using CHEM3D. The Raman, FTIR, and XPS spectra suggest that π-π interactions could take place between the CSH and IPF. Considering the pKa value of IPF (4.91) and pHiep of the CSH (3.21), electrostatic repulsion between the negatively-charged CSH and anionic IPF could play a negative role in the adsorption. A pore-filling mechanism could contribute to the adsorption in view of the molecular size of IPF (9.43 Å × 7.75 Å × 6.23 Å) and the average pore diameter of the CSH (2.27 nm). In addition, hydrophobic interactions could be involved in the adsorption. Multi-factor adsorption experiments were executed with pH, temperature, CSH dosage, and initial IPF concentrations as input variables and IPF removal rate as an output variable, and an artificial neural network (ANN) model with a topology of 49111 was developed to sufficiently describe the adsorption data (R > 0.99). Further analyses with additional experimental data confirm that the ANN model possessed good predictability for multi-factor adsorption.