Unveiling the biosorption performance of silk fibroin fiber for Pb(II) and Cr(VI) removal: Machine learning and DFT analysis

This study investigates the application of silk fibroin fiber (SFF) as an efficient biosorbent for lead (Pb(II)) and chromium (Cr(VI)) removal from water. Artificial neural network (ANN) and polynomial regression models (PRMs) were employed for prediction and assessment of batch adsorption tests, with their accuracy evaluated using various statistical measures. Under the varying operating conditions, optimized ANN predicted the maximum adsorption capacity (q_max) of ~39.69 μmoles g⁻¹ for Pb(II) at higher pH: 10.59. In contrast, Cr(VI) exhibited a higher q_max (140.38 μmoles g⁻¹) at elevated substrate concentrations and extended contact times. Statistical analysis revealed that higher-order PRMs achieved improved accuracy, with the mean squared error (MSE) decreasing to 92.05 %. While PRMs demonstrated competitive accuracy at higher error thresholds. Contact time emerged as the most crucial factor for Pb(II) adsorption (contributing 45.7 %), followed by temperature and pH. Cr(VI) adsorption exhibited a comparable trend. Breakthrough curves revealed an inverse relationship between influent concentration and breakthrough time. Higher influent concentrations led to faster saturation of adsorption sites, resulting in quicker breakthrough times for all models. Similarly, higher flow rates resulted in faster breakthrough due to the same effect. Density functional theory calculations suggested SFF's strong affinity for Cr(VI) compared to Pb(II) which involves a combination of Van der Waals and coordinate covalent bonds type mechanism, with preferential adsorption through carbonyl and amine groups. Notably, the metal uptake capacity remained consistent after regeneration with EDTA, signifying the successful reusability of the SFF bed, demonstrating SFF's potential as a robust reusable biosorbent for decontamination.