Machine Learning Guided Assembly of a Nested Gd₂₀@Gd₃₂ @Ni₃₆ Cluster via Urea Controlled Carbonate Release

Lanthanide-based polyhedral clusters are of great interest due to their unique geometries and functional properties, but their controlled synthesis remains a major challenge. Here, we report a nested three-shell cluster, [Gd₅₂Ni₃₆(MeIDA)₃₆(OH)₁₁₄(CO₃)₁₂ (H₂O)₄₈]·(ClO₄)₁₈·(H₂O)₂₁ (Gd₅₂Ni₃₆, H₂MeIDA = N-Methyliminodiacetic acid), achieved by using urea as a slow-release carbonate source to direct the formation of a dodecahedral inner shell. Single-crystal X-ray diffraction reveals a unique Gd₂₀@Gd₃₂@Ni₃₆ arrangement, with the innermost Gd₂₀ forming a Platonic dodecahedron templated by carbonate. A machine learning-guided, high-throughput synthesis platform enabled the exploration of 780 reaction conditions, uncovering key parameters and phase boundaries governing cluster formation. This work demonstrates data-driven strategies can accelerate the discovery of complex lanthanide architectures.