Glass composition affects properties through structure, and this composition→structure→property relationship is one of the core topics in material science. Glass property prediction and composition design based on the simulation models become an important research interest in glass development. Glass structure gene modeling (GSgM) is a practical method to introduce the measured glass structure into simulation in a real sense. This paper was to deal with the theoretical background, model establishment, recommendation for glass design method, structural data collection as well as the limitations of GSgM. The application of GSgM in property prediction, such as physical, chemical, thermal and spectroscopic properties, and glass composition design were demonstrated in the study of laser glass and solidification glass for simulated radioactive nuclear waste.

The theoretical basis of GSgM was firstly introduced, and then the effective glass design method for the collection of simulation dataset was recommended in detail. In addition, the composition-structure (C-S), structure-property (S-P) models and the bi-directional modeling of C↔S↔P in laser glass and radioactive waste solidification glass as well as the error correction function of this method were also established.

GSgM can ignore the complexity of the glass system, and all the measured glass structure data can be used for modeling. If the fine glass structure analysis (i.e., NMR or synchrotron radiation) is not available, the Raman and IR spectra are the most convenient and universal methods for glass structure characterization, in which the integrated area from the Gaussian peak fitting can be used as the structure information in the modeling. The peak-fitting is recommended to follow the characteristic vibration bands reported in references in order to decrease the relative error of the structure information. In the study of simulated radioactive waste solidification glass, The S-P modeling of chemical stability, thermal property, liquidus temperature and Mo-yellow phase models based on limited experimental data are performed to exhibit a high simulation accuracy of glass property prediction when the structure is used to do the simulation, and the model validation proves the reliability of the method. In Nd:phosphate laser glass, although the spectroscopic properties perform highly a non-linear relationship to glass composition, they can be still designed in a high accuracy via bi-directional simulation of C→S→P and P→S→C. There is another function of GSgM on error correction of experimental data. In the modeling of T_g and T_L, the results show that samples with a large property measurement error can be corrected by GSgM simulation.

Glass structure gene modeling (GSgM) was a simulation method in which glass structure was used as a “bridge” to thread composition and property by C-S and S-P modeling. It could provide an effective prediction and optimization method for glass simulation. The high accuracy of GSgM could come from the transformation of the nonlinear relationship of C-P to two linear relationships of C-S and S-P. The GSgM approach was demonstrated as an alternative, powerful simulation method for glass modeling with limited data, and the relationship between glass composition, structure and properties could be further understood so as to achieve the more efficient simulation of glass design and glass property.