Machine learning-guided design of novel glasses with good vitrification ability and desired thermal expansion coefficient

• Grant number: 23/13961-0
• Support Opportunities: Scholarships in Brazil - Program to Stimulate Scientific Vocations
• Start date: January 04, 2024
• End date: February 23, 2024
• Field of knowledge: Engineering - Materials and Metallurgical Engineering - Nonmetallic Materials
• Principal Investigator: Edgar Dutra Zanotto
• Grantee: Káliton Pereira Mororó
• Host Institution: Centro de Ciências Exatas e de Tecnologia (CCET). Universidade Federal de São Carlos (UFSCAR). São Carlos , SP, Brazil

Abstract

In this project we intend to: 1) to use a computer code that combines several data-driven predictiveMachine Learning models (Random Forest, Artificial Neural Networks and others using stackedgeneralization) coupled with a Genetic Algorithm to design novel oxide glass formulations havingdesired thermal expansion coefficient. The computer code will employ combinations of elementproperties that could also predict the vitrifiability tendency or glass forming ability (GFA) of the suggested formulations, such as, for instance: electronegativity and Shannon Entropy; 2) produce the glasses in the laboratory; and 3) measure their thermal expansion coefficient and compare with thepredictions. The first part has been pioneered by our group 5 years ago and several teams in the world are now working on this strategy; however, to the best of our knowledge, nobody has used the stakingprocedure for this glass design problem. Also, the second part is original at international level; so far there is no published report on the GFA prediction of ML-derived oxide glass compositions. This soundslike a one-year proposal; however, a student in LaMaV (Rafael B. Rosante) has been working with usfor several months already and has great ability with coding and ML algorithms to assist in the first part. (AU)