Abstract
Understanding how local structural units quantitatively govern transport properties is a central issue in high-temperature aluminate melts. In this work, binary (100-x)CaO-xAl2O3 melts (x = 25, 31, 37, 43, and 50 mol%) were investigated at 1923 K by combining in situ high-temperature Raman spectroscopy with classical molecular dynamics (CMD) simulations, enabling a quantitative analysis of local structure and its correlation with transport properties. The melt network is dominated by [AlO4] tetrahedra over the investigated composition range. With the increasing Al2O3 content, the fraction of higher-coordinated aluminum species ([AlO5] and [AlO6]) increases markedly, with their combined population rising from 3.81 to 14.01%. This increase is accompanied by a concurrent enhancement of edge-sharing Al-O polyhedral linkages, indicating a consistent evolution of coordination environment and network connectivity. Raman spectra corrected for reduced Raman scattering cross sections enable reliable quantification of Qi species, revealing a progressive transformation from low-polymerized Q0-Q2 units toward a highly polymerized network dominated by Q3 and Q4 species, consistent with CMD-derived structural trends. With the increasing network polymerization, the self-diffusion coefficients of Ca2+, O2−, and Al3+ decrease systematically, while melt viscosity increases and electrical conductivity decreases. Based on the quantitatively determined Qi species, a structure-based viscosity model was established, in which the activation energy for viscous flow is expressed as a weighted function of Qi species. The relative contributions to viscosity follow the order Q4 > Q3 > Q2 > Q1, providing a predictive link between local structure and macroscopic viscosity in the melts.