Not random at all: modifier preferences and local environments in polyphosphate glasses. Insights from thermodynamic modeling and MAS NMR measurements

• Network modifier preferences: potassium shows a strong preference for coordinating with Q 1 units, while sodium predominantly associates with Q 2 units; calcium increasingly favors Q 2 environments as potassium substitutes sodium. • Sodium environment evolution: Experimental 23 Na MAS-NMR data aligns with model predictions, showing systematic shifts in chemical shifts due to changes in sodium coordination number and local electron density. • Qn distribution accuracy: The thermodynamic model reliably predicts the Q n distribution, with good agreement when compared to 31 P MAS-NMR measurements. • Q 2 vs Q 1 behavior: The model explains deviations in experimental Q 2 chemical shifts effectively, although it does not fully resolve the behavior of Q 1 environments. Structural studies of glasses have traditionally focused on the connectivity of Qⁿ tetrahedra forming the glass network. However, this approach provides limited insight into the atomic-scale distribution of network modifiers (NWM). Our aim was to investigate their spatial distribution in K₂O/Na₂O–CaO–MgO–P₂O₅ polyphosphate glasses, using thermodynamic modeling and comparing its results with findings from 31 P and 23 Na MAS NMR measurements. Results reveal that modifiers show distinct preferences for certain local environments: potassium strongly prefers Q¹ units, while sodium predominantly associates with Q². Additionally, calcium progressively shows a preference for oxygens associated with Q² rather than Q¹ units as potassium for sodium substitution increases. Comparing these results with experimental 23 Na MAS NMR data, we find that they provide a consistent interpretation of the evolving local sodium environment during this substitution. Variations in sodium coordination number, the electronegativity of its next-nearest neighbors, and changes in electron density explain the systematic changes observed in the 23 Na chemical shift. The model accurately predicts the Qⁿ distribution compared with 31 P MAS NMR data and offers a plausible explanation for deviations in the experimental Q² chemical shifts, though the behavior of Q¹ remains unresolved. A further limitation of the model affects the calculation of molar volume, owing to its initial assumptions not accounting for free volume effects, with relative errors ranging from 1% to 12% compared to experimental data.