Beyond Conversion: Network Topology and Complementary Structure-Property Relationships during Epoxy Curing

Predicting thermoset properties from the evolving molecular network remains challenging, because chemical conversion alone does not capture the emergence of mechanically relevant topology. Here, molecular dynamics simulations (~100,000 atoms) of EPON 862/DETDA curing are combined with graph-theoretical network analysis across eleven conversion levels up to 94%. Minimum-cycle-basis tracking and giant-component decomposition show that gelation marks the onset, not the completion, of topological densification: the biconnected core grows from 3.1% of all network nodes at the gel point to 88.9% at near-complete cure, rising most steeply between 60% and 70% conversion, where separate cores coalesce into a single dominant backbone. Univariate regression with AICc model selection shows that descriptor validity is regime-dependent: after gelation, network topology describes large-deformation and failure behavior better than conversion, whereas thermal and mobility properties gain nothing from topological descriptors. Descriptor selection for network-forming materials should therefore follow the regime of network formation.