Use of adsorption-influencing parameters for designing the batch adsorber and neural network-based prediction modelling for the aqueous arsenate removal using combustion synthesised nano-alumina.

Removal of arsenic from water is of utmost priorities on a global scenario due to its ill effects. Therefore, in the present study, aluminium oxide nano-particles (nano-alumina) were synthesised via solution combustion method, which is self-propagating and eco-friendly in nature. Synthesised nano-alumina was further employed for arsenate removal from water. Usually, pre-oxidation of arsenite is performed for better removal of arsenic in its pentavalent form. Thus, arsenate removal as a function of influencing parameters such as initial concentration, dose, pH, temperature, and competing anions was the prime objective of the present study. The speciation analysis showed that H2AsO4- and HAsO42- were co-existing anions between pH 6 and 8, as a result of which higher removal was observed. Freundlich isotherm model was well suited for data on adsorption. At optimal temperature of 298 K, maximum monolayer adsorption capacity was found as 1401.90 μg/g. The kinetic data showed film diffusion step was the controlling mechanism. In addition, competing anions like nitrate, bicarbonate, and chloride had no major effect on arsenate removal efficiency, while phosphate and sulphate significantly reduced the removal efficiency. The negative values of thermodynamic parameters ΔH° (- 23.15 kJ/mol) established the exothermic nature of adsorption, whereas the negative values of ΔG° (- 7.05, - 6.51, - 5.97, and - 5.43 kJ/mol at 298, 308, 318, and 328 K respectively) indicated the spontaneous nature of the process. The best-fitted isotherm was used to design a batch adsorber to estimate the required amount of aluminium oxide nano-particles for achieving the desired equilibrium arsenate concentration. Nano-alumina was also applied to treat the collected arsenic-contaminated groundwater from actual field. Experimental data were used to develop a neural network-based model for the effective prediction of removal efficiency without carrying out any extra experimentation.