Identifying silica types using viscosity data and principal component analysis

Abstract

As the most important glass former, silica has a plethora of relevant industrial and commercial applications. Silica glass can be classified into several types depending on the starting materials, production method, content and impurities. Briefly, type I is obtained by melting natural or synthetic quartz in electrical furnaces, presenting alkali and metal impurities (in ppm). Type II is obtained from melting natural or synthetic quartz in hydrogen–oxygen or natural gas flames with few impurities (in ppm) but high concentration of structural water. Type III is obtained from high-temperature hydrolysis of volatile compounds of silicon with low metal impurities and considerable concentrations of structural water and chlorine. Type IV is obtained from high-temperature oxidation of silane with small amounts of metal impurities, almost no structural water but high chlorine content. We analyzed literature data on viscosity, η, for 44 batches over a wide temperature range, from glass transition, Tg, to far beyond the melting point, Tm, for four of five silica types. The viscosity behavior in silica follows the Arrhenius expression: log10η(T)=A+BT, with η in Pa⋅s, T in K, where A and B are constants. However, the Vogel-Fulcher-Tammann-Hesse (VFTH) equation: log10η(T)=A+BT−T0,can also be applied to this system, with T0 → 0 as another constant (in K). For a fixed temperature, a variance of η, covering two orders of magnitude, was observed for types I and II, and about one order for the other two types, mainly due to impurity content and the manufacturing process. Another important observation was that the highest viscosities were found in the purest silicas. Not all samples presented T0 = 0, as expected for Arrhenius viscosity behavior. An analysis considering the Mauro-Yue-Ellison-Gupta-Allan (MYEGA) viscosity equation was also performed. Previous studies of principal component analysis (PCA) considering A, B and T0 allowed the mapping of binary and ternary silicate and borate systems. In brief, PCA is a mathematical technique which rearranges information from datasets and can classify silica types according to their viscosity parameters. The results show that it is possible to map the four silica glass types according to viscosity VFTH and MYEGA parameters taken from experimental data.