Predicting adsorption capacity of pharmaceuticals and personal care products on long-term aged microplastics using machine learning

Jingjing Yao,Haipu Li,Hui Ying Yang

DOI: https://doi.org/10.1016/j.jhazmat.2023.131963

IF: 13.6

2023-07-04

Journal of Hazardous Materials

Abstract:We investigated the adsorption mechanism of 66 coexisting pharmaceuticals and personal care products (PPCPs) on microplastics treated with potassium persulfate, potassium hydroxide , and Fenton reagent for 54, 110, and 500 days. The total adsorption capacity (q e ) of 66 PPCPs on 15 original microplastics was 171.8 – 1043.7 μg/g, far below that of 177 long-term aged microplastics (7114.0 – 13,114.4 μg/g). Around 69.8% of q e was primarily influenced by the total energy, energy of the highest occupied molecular orbital, and energy gap of PPCPs, calculated using the B3LYP/6–31 G* level. Furthermore, 111 aged microplastics exhibited similar total q e values. Additionally, we developed predictive models based on attenuated total reflectance Fourier transform infrared spectroscopy to predict the individual and total q e on 192 microplastics. These models, including the maximal information coefficient and gradient boosting decision tree regression, exhibited high accuracy with R training 2 values of 0.9772 and 0.9661, respectively, and p-values below 0.001. Spectroscopic analysis and machine learning models highlighted surface functional group alterations and the importance of the 1528–1700 cm −1 spectral region and carbon skeleton in the adsorption process. In summary, our findings contribute to understanding the adsorption of PPCPs on microplastics, particularly in the context of long-term aging effects.

environmental sciences,engineering, environmental