Topological descriptors for disordered networks enable property prediction of multicomponent oxide glasses

Establishing the structure-property relationships in multicomponent oxide glasses is highly challenging due to their inherently compositional and structural complexity. In this work, we present a topology-inspired structural decoding strategy to predict the structural impact on the properties of the 45S5-based mixed alkali bioactive (MAB) glasses, a prototypical multicomponent system. Two novel topological descriptors, angular rigidity ( γ ac ) and persistent homology matrix ( P H m ), are introduced to map the structure-property landscape in MAB glasses. In detail, γ ac accounts for the spatial averaging of network distortions, quantitatively predicting properties associated with network reorganization, such as glass transition temperature and hardness. P H m evaluates the topological similarity between glass and crystalline phases, thereby characterizing the resistance of MAB glass against crystallization (i.e., glass stability). Compared with conventional network analysis approaches, our two topological descriptors concurrently capture the higher-order nonlinear evolution of properties driven by the synergistic interplay between composition and cooling history. Importantly, these topological descriptors are derived solely from static glass structures generated via experimental measurements, suggesting potential utility for understanding diverse disordered materials. This work thus establishes a closed-loop framework that integrates synthesis, characterization, and predictive modelling for the specific 45S5-based MAB glass systems. The demonstrated ability to correlate topological features with distinct properties within this system represents a step towards the rational design of multicomponent oxide glasses.