The relative binding energies of lanthanide–EDTA complexes, at varying solution pH, were predicted from electronic structure calculations using the solution structures of Ln–EDTA complexes resolved withab initiomolecular dynamics simulations.

The lanthanide elements are crucial components in numerous technologies, yet their industrial production through liquid-liquid extraction continues to be economically and environmentally costly due to the challenge of separating elements with similar physicochemical properties. While computational ligand screening has shown promise toward discovering efficient extractants, the complexity of constructing chemically sensible 3D structures (often by hand), coupled with the high cost of quantum chemistry calculations, often limits exploration of the vast ligand chemical and conformational space in favor of local exploration around known chemistries. Moreover, metal complexes can have many stable configurations whose differences in energies exceed the small energy differences that determine the extractant selectivity for certain lanthanides. Because of this difference, incorrect selectivity predictions can be made if the lowest energy coordination complex is not identified and modeled. To address this issue, we present a high-throughput computational workflow that automates the construction and quantum mechanical modeling of 3D lanthanide-extractant complexes. This approach allows for an unbiased search of distinct configurational and compositional variations for each metal, enabling accurate predictions of their solution structures and lanthanide selectivity. As showcased by three extractants from diverse chemical categories-a crown ether, a phenanthroline monocarboxamide, and a malonamide-it is found that sampling the lanthanide-ligand configuration space is critical to correctly predicting the metal coordination environment and experimental lanthanide selectivity trends.

Most QM-cluster models of enzymes are constructed based on X-ray crystal structures, which limits comparison to in vivo structure and mechanism. This work introduces an MD to QM-cluster model workflow.