Tetracycline sorption by magnetic biochar derived from watermelon rind: performance and influential factors

Abstract

Tetracycline (TC) antibiotic is one of emerging contaminants in water reservoirs that causes undesirable effects on environment and human health. Magnetic biochar (MBC) is considered a promising sorbent in adsorption process for removal of contaminants with highly efficient and facile operation. In this work, MBC was synthesized by pyrolysis of watermelon rind impregnated with FeCl3 at different pyrolysis temperatures in a range of 600-900 °C prior to applying for TC adsorption. Characteristics of MBC were analyzed by scanning electron microscopy, elemental analyzer, N2 adsorption/desorption, Fourier-transform infrared spectroscopy, Raman spectroscopy, vibrating sample magnetometry, and X-ray diffractometry. The adsorption kinetics, isotherm, effect of solution pH, and reusability were investigated. Moreover, an emperical and semi-empiriacl model of TC adsorption capacity under influential factors based on response surface methodology (RSM) and machine learning (ML) were developed. From the results, an increase in the pyrolysis temperature from 600 to 900 °C significantly affects on characteristics of MBC. The adsorption kinetics of MBC600 follows pseudo-first-order kinetic model while MBC700, MBC800, and MBC900 follow pseudo-second-order kinetic model. The adsorption isotherm fitted well with Freundlich isotherm investigating heterogeneous adsorption site. In addition, the highest maximum adsorption capacity of 77.60 mg/g could be obtained from MBC900. The adsorption process is pH-dependent. The reusability test revealed adsorption capacity decrease from 100 to 83.89% after five cycles. ANOVA results confirmed an empirical model was significantly at 95% confidence. In addition, the order of influential terms is solution pH > interaction between initial TC concentration and solution pH > square effect of initial TC concentration > initial TC concentration. Three different ML algorithms were used to develop the semi-empirical model. It was found that Catboost outperformed random forest and boosted regression tree. The feature important revealed SBET provided the largest effect on TC adsorption capacity followed by (O+N)/C, initial TC concentration, H/C, and C%, respectively.