Abstract
The removal of antibiotics from aqueous media along with their recovery is still an open research topic, due to their practical and economical importance. Adsorption allows these two objectives to be achieved, provided that the adsorbent used is chemically and mechanically stable and has a low preparation cost. In this study, PET (polyethylene terephthalate) fibers, obtained by mechanically processing PET waste, were used for the adsorption of rifampicin (RIF) and rifaximin (RIX) antibiotics from aqueous media. The experimental adsorption capacity of PET fibers for the two antibiotics (RIF and RIX) was determined at different pH values (2.0-6.5), adsorbent dose (0.4-20.0 g/L), contact time (5-1440 min), initial antibiotic concentration (4.0-67.0 mg/L), and temperature (10, 22, and 50 °C); the experimental values of these parameters were analyzed using a neuro-evolutive technique (ANE) combining sequential deep learning (DL) models with a differential evolution algorithm. The obtained optimal ANN-DL algorithm was then used to obtain the optimal models for the adsorption of RIF and RIX on PET fibers, which should adequately describe the adsorption dynamics for both antibiotics. The adsorption processes are spontaneous and endothermic (ΔG < 0, ΔH > 0) and are described by the Langmuir model (R(2) > 0.97) and the pseudo-second order kinetic model (R(2) > 0.99). The retention of RIF and RIX on the surface of PET fibers occurs through physicochemical interactions, and the FTIR spectra and microscopic images support this hypothesis. The presence of inorganic anions in the aqueous solution leads to an increase in the adsorption capacities of RIF (max. 7.6 mg/g) and RIX (max. 3.6 mg/g) on PET fibers, which is mainly due to the ordering of water molecules in the solution. The experimental results presented in this study allowed for the development of the adsorption mechanism of RIF and RIX on PET fibers, highlighting the potential practical applications of these adsorption processes.