Rheological properties of SiO2-CaO-MgO-Al2O3-(0-30 wt%)B2O3 system at basicity 1.4: Experiment and modelling

Abstract

Boron oxide (B₂O₃) affects the viscosity and phase change temperatures of glasses and melts, which are widely used in various industrial processes. In this study, the effect of B₂O₃ in a wide concentration range (0–30 wt%) on the melting and rheological characteristics of the oxide system SiO₂(37.7–25.2 wt%)-CaO(42.8–25.3 wt%)-MgO(10 wt%)-Al₂O₃(9.5 wt%)-B₂O₃(0–30 wt%) with a constant basicity of 1.4 was studied experimentally and theoretically. The experimental rheological data measured with a high-temperature rheometer up to 1550 °C were supported by the study of the internal structure by an X-ray powder diffraction analysis (XRPD), scanning electron microscopy with an energy-dispersive X-ray spectroscopy (SEM/EDX), a Fourier transform infrared spectroscopy (FTIR) and Raman microspectroscopy. The increase in the B₂O₃ content in the samples caused a dramatic decrease in the softening and liquidus temperatures. Up to 15 wt% B₂O₃ content, it was possible to determine the break temperature, which decreased with increasing B₂O₃ content. The exponential decrease in dynamic viscosity with an increasing temperature and its decrease with an increasing B₂O₃ content was caused by the decrease in the [BO₄] tetrahedra coupled with an increase in the [BO₃] units and boroxol rings, leading to a simplified structure with a higher number of non-bridging oxygens. Semiempirical models (Arrhenius, VFT, Riboud, Urbain, and Ray and Pal) were used to compare experimentally acquired dependencies with theoretically obtained ones. The dependence of dynamic viscosity over the temperature range of 1320–1520 °C and a B₂O₃ content 0–30 wt% was predicted with a relative error of 2.5 % using an artificial neural network (ANN). Unlike the empirical models, the ANN is a versatile statistical tool which can predict the dynamic viscosity of melts without assuming the physical nature of the processes.