Harnessing Dynamic Supramolecular Interactions for Lanthanide Detection via Computational Pattern Recognition of Magnetic Resonance Fingerprints

The reliance of modern technology growth on lanthanides presents dual challenges: securing sustainable sources from natural or recycled materials and reducing environmental harm from waste discharge. However, the similar ionic radii, oxidation states, and binding affinities of Ln3+ ions hinder their nondestructive detection in mixtures. Furthermore, the overlap of spectroscopic signals and the inapplicability for opaque solutions limit the harness of luminescent sensors for differentiating one Ln3+ from another. Here, we introduce 19F-paramagnetic guest exchange saturation transfer magnetic resonance fingerprinting (19F-paraGEST MRF), a rapid signal acquisition, encoding, and analysis approach for detecting specific Ln3+ in mixtures. Based on a small-sized experimental 19F-paraGEST data set, we generated a de novo dictionary of ∼2500 combinations of Ln3+ mixtures, resulting in ∼7,000,000 simulated 19F-paraGEST MRF patterns of different Ln3+ concentrations. This dictionary was later used for computational pattern recognition of experimental NMR signal evolutions (“fingerprints”), utilizing a rapid computational approach executable on a standard laptop within seconds. Hence, fast and reliable multiplexed lanthanide detection in complex mixtures was enabled. Demonstrated through the analysis of lanthanides’ content of permanent magnets from a hard disk drive, this MR-based method paves the way for broader applications of lanthanide detection in murky, nontransparent mixtures and further exploration of supramolecular sensors in diverse scenarios.