Attempts to scientifically capture and quantify glass homogeneity have been based on optical methods, namely, on those based on the work of Christiansen, Raman, and Shelyubskii. In industrial practice, a number of further methods are in use. In this study, glass homogeneity is determined by the measurement of density fluctuation. The method of "density titration" proposed in earlier literature is used in an essentially improved procedure. In this method, the density distribution is determined for a representative ensemble of glass particles prepared from a bulk glass sample. The particles float on a heavy liquid which is diluted, step by step, by a miscible liquid of lower density. The particle fraction with density exceeding the density of the mixture sinks. When the method is performed on quenched or hyper-quenched glass, then two types of inhomogeneities contribute to the overall homogeneity, i.e. thermal and chemical contributions. By two consecutive measurements, one performed on a sample of quenched glass, one on the same sample relaxed at T = 0.9Tg, these contributions are clearly discriminated. Results are presented by an error function type distribution where the median marks the average density, and the standard deviation is adopted as a measure of glass homogeneity.
