Abstract:

A statistical modeling approach to modeling glass composition (C) - structure (S) - property (P) is introduced based on glass property response to the glass network structure. This paper first reviewed some of the limitations of the C-P statistical modeling approach, then followed by complementary benefit identified from using S-P statistical modeling approach. Furthermore, S-P modeling is not limited by a narrower composition space as seen in the C-P modeling case, which benefits glass composition fine-tuning and design optimization, such as in the chemical stability experiment for Nd: phosphate laser glass, the S-P models perform much better than the C-P models. The procedure of C-S-P modeling was illustrated, and how to use C-S and S-P models inverse the composition of glass was also detailed. Except for the regular properties, C-S-P modeling methodology can provide more accurate predictions on laser glass emission properties, chemical durability, etc., which are often difficult by using the C-P modeling approach alone. Our effort on C-S-P modeling is to explore a general methodology that can provide researchers with an alternative method to facilitate glass design with higher efficiency, fast turn-around, and high accuracy and precision.

Key words: glass structure gene, statistical analysis modeling, composition-structure-property model