Machine-Learning-Guided Identification of Coordination Polymer Ligands for Crystallizing Separation of Cs/Sr

Area of Science:

• Coordination chemistry
• Materials science
• Nuclear energy

Background:

• Spent nuclear fuel reprocessing is crucial for sustainable nuclear energy.
• Efficient separation of Cesium (Cs) and Strontium (Sr) is a key challenge.
• Crystallization-based separation requires effective ligand design.

Purpose of the Study:

• To develop a machine learning (ML)-guided methodology for screening ligands for selective Strontium (Sr) binding.
• To identify and experimentally validate ligands for Sr/Cs separation via crystallization.
• To advance nuclear fuel reprocessing technologies.

Main Methods:

• Mining the Cambridge Structural Database for coordination chemistry data.
• Developing and optimizing an ML model to predict ligand-metal binding properties.
• Experimental validation of ML-identified ligands (squaric acid, chloranilic acid) for Sr/Cs separation.

Main Results:

• An optimized ML model successfully ranked potential ligands for Cs/Sr selective crystallization.
• Squaric acid and chloranilic acid demonstrated high Sr²⁺ sequestration capability.
• These ligands exhibited significant selectivity for Sr²⁺ over Cs⁺.

Conclusions:

• The AI-guided strategy effectively identified promising ligands for Sr/Cs separation.
• Experimental results validated the ML model's predictions.
• This approach accelerates the development of advanced nuclear fuel reprocessing techniques.