Citric acid-modified loquat seed waste as biosorbent for methylene blue: kinetics, isotherm, thermodynamics, and machine learning-based prediction of adsorption performance

Abstract

ABSTRACT Graphical abstract showing citric-acid modification of loquat seed powder and its application as a biosorbent for methylene blue removal. The material achieved a maximum adsorption capacity of 47.53 mg/g within 9 min, followed by the Freundlich and pseudo-second-order models, which were accurately predicted by a random forest model, and retained more than 85% removal efficiency after five reuse cycles. Synthetic dye contamination poses a major environmental challenge, requiring sustainable, high-performance remediation technologies. Herein, an eco-friendly biosorbent, citric acid-modified loquat powder (LP-CA), was developed for efficient removal of methylene blue (MB) from aqueous solutions. FTIR, FE-SEM-EDS, and pHpzc analyses confirmed successful surface functionalization by introducing oxygen-containing groups, enhancing surface reactivity and yielding a pHpzc of 4.0. LP-CA reached adsorption equilibrium within 9 min and achieved a maximum adsorption capacity of 47.53 mg/g, representing a 19% improvement over the pristine biomass. Equilibrium and kinetic data were best described by the Freundlich isotherm (R2 = 0.98) and the pseudo-second-order model. Thermodynamic and activation energy analyses revealed a spontaneous, exothermic process predominantly governed by physisorption, with electrostatic attraction, hydrogen bonding, and π-interactions contributing to adsorption. A Random Forest model accurately predicted MB removal (R2 = 0.970; RMSE = 3.57 %), and feature importance analysis identified solution pH and pHpzc as the dominant variables, quantitatively validating the proposed adsorption mechanism. LP-CA retained >85% removal efficiency through the first five adsorption-desorption cycles and demonstrated practical applicability in a bag-assisted adsorption configuration. Machine learning provides a powerful framework for predictive adsorption modeling, mechanistic interpretation, and future process optimization.