Context-Guided Tree Similarity for Compound-Derived Structural Comparison of Rare-Earth Elements

Source abstract

Quantifying structural similarity among chemical elements is important for understanding compositional relationships in chemical systems. In this work, we introduce a compound-derived growth-tree representation that characterizes each element through its compositional expansion patterns across reported compounds. Based on this representation, we propose a Context-Guided Tree Similarity (CGTS) measure, which evaluates similarity by aggregating local overlaps under shared compositional contexts rather than relying solely on global comparisons. Applying this framework to rare-earth elements, we obtain a distinctly non-uniform similarity structure, with high similarity concentrated among heavy rare-earth elements such as Tb, Dy, Ho, Er, Tm, and Lu. Analysis of top-ranked pairs reveals strong consistency in local expansion behavior within shared chemical environments. These results suggest that compound-derived structural similarity provides a meaningful basis for analyzing relationships among chemical elements from a compositional perspective. The identified similarity patterns are consistent with known chemical affinities within the rare-earth series, indicating that the proposed framework captures structurally relevant chemical behavior. The approach therefore provides a compound-derived descriptor for comparing elemental similarity in chemical systems.