Data-driven exploration of layered double hydroxide crystals exhibiting high fluoride ion adsorption properties and chemical stability

Abstract

Fluoride ion (F⁻) contamination of groundwater has become a global issue. As potential adsorbents for F⁻ removal, layered double hydroxides (LDHs) have moderate affinities for F⁻. Moreover, the preparation of LDHs exhibiting both high F⁻ adsorption capacities and chemical stability is empirically challenging. To overcome this issue, we used process informatics to explore promising ternary LDHs with high F⁻ adsorption capacities and chemical stability. We constructed machine learning models based on F⁻ adsorption test data and Bayesian optimisation. Initially, the objective variable for LDH candidates was the F⁻ adsorption amount. By considering LDH systems that combine one type of divalent cation (M²⁺) with two types of trivalent cation (M³⁺), ternary LDHs such as Ni–Fe–Ga and Ni–Al–Ga LDHs, which have not been studied previously, were proposed. The subsequent addition of the M²⁺ leaching amount as an objective variable allowed the identification of LDHs such as Ni–Fe–Y and Ni–Cr–Y LDHs with high F⁻ adsorption capacities (15–17 mg g⁻¹ at 1 mM F⁻, K_d > 4600–8300 mL g⁻¹) and chemical stability. Projected crystal orbital Hamilton population analysis indicated that the M²⁺–O bonds in Ni–Al–Ga and Ni–Cr–Y LDHs have a stronger covalent character than those in Mg-based LDHs. These findings provide guidelines for the synthesis of novel LDHs with various compositions.