Statistical glass structure gene modeling on liquidus temperature of high level waste glasses

Abstract

Statistical glass structure gene modeling (GSgM) on liquidus temperature (TL) of simulated high level waste (HLW) glass is detailed in this article. TL values of a set of HLW borosilicate glasses were experimentally determined by x-ray diffraction (XRD) method with 24 h heat-treated glass in gradient temperature furnace, and the primary crystalline phase was CaMoO4. Integral peak areas obtained from the FTIR peak fitting were used as the structural data, and the statistical TL-structure model was first established. The S-P (TL) model was shown to have a high accuracy, i.e., R2 0.978, P value 0.0009 and RMSE 7.6 °C. Model validation showed the predicted and measured TL values were 960.4 °C and 965 °C, respectively. Model iteration further reduced P value to 0.0001 and RMSE to 7 °C. Relationship between composition (C) and the glass network structure (S) was then simulated, and composition-FTIR structure-TL analysis indicated that SiO2, ZnO, rB3+, MoO3, ZrO2, B2O3 and rare earth affect TL by influence the structural units Si–O–Si rocking vibrations, νs (Si–O–Si) in [SiO4] and νs ([BO3] or BO2O−). Results showed that GSgM offers an alternative approach to simulate TL of complex glass with limited experimental data, providing a plausible linkage between composition, structure and property.