Statistical approach to modeling relationships of composition - structure - property I: Alkaline earth phosphate glasses

This paper introduces a methodology of accurate property predictions for amorphous materials (focusing on oxide phosphate glass) based on statistical composition (C) - structure (S) - property (P) modeling. Utilizing literature reported 31P MAS NMR data (Qn, δiso, η) of binary alkaline earth phosphate glasses RO-P2O5 (R = Mg, Ca, Sr, Ba), unified relationships of RO and Qn (n = 2, 3) were firstly established. A statistical approach to modeling mathematic relationships of C-S, S-P and then C-S-P for the alkaline earth binary phosphate glasses was illustrated step by step. The C-S models of Q2 and Q3 were then used to calculate Q2 and Q3 in the mixed alkaline earth (MAE) phosphate glass, MgO-CaO-P2O5, which in turn result in S-P models accurately tracks the glass properties, density, Young's modulus, glass transition temperature, and Microhardness. The S-P models were also used to simulate non-linear relationships of the MAE phosphate glasses as a function of Q2/(Q2+Q3) from the glass network structure view point in comparison with the property responses to the change of MgO/(MgO + CaO). Within the entire composition space of the alkaline earth phosphate glasses with RO/P2O5 ≤ 1, the statistical C-S-P modeling for predicting glass properties, including glass transition temperature, Young's modulus, density, etc., requires no assumptions on glass structures or other constraints as compared with other theoretically based modeling approaches.