ADF helps guide rare-earth separations discovery

Rare-earth separations are difficult because neighboring lanthanides have similar ionic radii and usually occur as +3 ions. In a recent JACS paper, Augustine, Wang, Taylor, Tynes, Adelman, Batista, Kozimor, Lubbers, Perez, and Yang developed a data-driven workflow for separating Nd, Eu, Dy, and Ho using di(2-ethylhexyl)phosphoric acid (HDEHP) together with an aqueous holdback agent.

ADF in the Amsterdam Modeling Suite was used for the computational part of the workflow. After screening the IUPAC stability constants database, the authors selected five candidate holdback agents and generated lanthanide coordination structures. More than 6000 combined Architector and ADF calculations were run, with ADF 2022.103 used for DFT refinement of the complexes. The calculations used PBE/TZP, ZORA, DFT-D4, and COSMO solvation models for the aqueous and organic phases.

The DFT calculations helped compare relative binding energies across the lanthanide series and guided the experimental down-selection. Automated solvent-extraction experiments then identified oxaloacetic acid as the best holdback agent in combination with HDEHP. At pH around 2.0, Eu, Dy, and Ho were extracted over Nd. At pH around 0.5, selectivity shifted toward separating Dy and Ho from Eu and Nd.

The study is a good example of where quantum chemistry fits in an autonomous discovery workflow. ADF did not replace experiment: third-phase formation ruled out some candidates and had to be observed experimentally. Its role was to provide a molecular filter and later help connect the observed pH dependence to oxaloacetic-acid protonation. Together with Bayesian optimization, the workflow found selective regimes after testing 180 unique conditions, rather than the 1280 conditions required for a full grid.