Iron phosphate glass waste forms for vitrifying Hanford AZ102 low activity waste (LAW), part I: Glass formation model

Abstract

A methodology for determining glass formation in a 5-component iron phosphate base glass system that contained P2O5, Fe2O3, Al2O3, Na2O and SO3 has been developed using a standard 5-component D-Optimal Statistical Mixture Design procedure. This compositional system was selected in order to study the feasibility of vitrifying a high soda (~80wt%), a high sulfate (~17wt%) containing Hanford AZ102 LAW in iron phosphate glasses with waste loadings much higher than the currently reported values. The statistical analysis was designed to develop a series of initial compositions (total 28), nearly half of which formed glass upon melting and cooling. After separating the compositions that formed glass from those that crystallized, three molar ratios, namely, (O/Na), (Na/Al) and (Na/P), were identified as potential parameters for defining glass formation (or crystallization). The (O/Na) molar ratio had the greatest effect. Melts whose (O/Na), (Na/P) or (Na/Al) molar ratios were ≤4.35, ≥1.13, or ≥2.56, respectively, had the highest likelihood to form glass when cooled. The melts crystallized completely when the values of these three molar ratios were, respectively, ≥6.02, ≤0.77 or ≤1.92. For the compositions that formed glass, the waste loadings as defined by the sum of Na2O and SO3, the two major components comprising ˃90wt% of the Hanford AZ102 LAW, ranged from 25 to 33wt%.A glass formation-composition model that formalized the above findings was developed using the Classification and Regression Tree (CART) analyses. The CART model was validated by testing the glass formation characteristics of several additional compositions whose (O/Na) molar ratio was maintained ≤4.35, i.e., the compositions were expected to form glass. This model is expected to identify similar molar ratios for predicting glass formation/crystallization of iron phosphate based compositions containing not only other Hanford LAW streams, but high level wastes as well. Using the compositions that formed glass, property-composition statistical models for the waste forms were also developed and will be reported in a second paper.