A Review of Fluoride Removal from Phosphorous Gypsum: A Quantitative Analysis via a Machine Learning Approach.

This review comprehensively explores fluoride removal from phosphogypsum, focusing on its composition, fluorine-containing compounds, characterization methods, and defluorination techniques. It initially outlines the elemental composition of phosphogypsum prevalent in major production regions and infers the presence of fluorine compounds based on these constituents. The study highlights X-ray photoelectron spectroscopy (XPS) as a pivotal method for characterizing fluorine compounds, emphasizing its capability to determine precise binding energies essential for identifying various fluorine species. Additionally, the first-principle density functional theory (DFT) is employed to estimate binding energies of different fluorine-containing compounds. Significant correlations are observed between the total atomic energy of binary fluorides (e.g., of alkali metals, earth metals, and boron group metals) and XPS binding energies. However, for complex compounds like calcium fluorophosphate, correlations with the calculated average atomic total energy are less direct. The review categorizes defluorination methods applied to phosphogypsum as physical, chemical, thermal, and thermal-combined processes, respectively. It introduces neural network machine learning (ML) technology to quantitatively analyze and optimize reported defluorination strategies. Simulation results indicate potential optimizations based on quantitative analyses of process conditions reported in the literature. This review provides a systematic approach to understanding the phosphogypsum composition, fluorine speciation, analytical methodologies, and effective defluorination strategies. The attempts of adopting DFT simulation and quantitative analysis using ML in optimization underscore its potential and feasibility in advancing the industrial phosphogypsum defluorination process.