Event Type:
MSE Grad Presentation
Date
Talk Title:
Design of (De)Polymerizable Polymers Using Machine Learning-Based Predictive Models and Generative Algorithms
Location:
Love 210 and via Microsoft Teams
https://teams.microsoft.com/l/meetup-join/19%3ameeting_OWViOTk3NmItMTU3YS00ZTRk…

Committee Members:
Dr. Rampi Ramprasad, Advisor, MSE
Dr. Karl Jacob, MSE
Dr. Sunderasan Jayaraman, MSE
Dr. Blair Brettmann, ChBE
Dr. Chao Zhang,  CSE

Design of (De)Polymerizable Polymers Using Machine Learning-Based Predictive Models and Generative Algorithms

Abstract:

Plastics are an important asset to humanity, particularly for elongating the shelf life of  food and preventing bacterial contamination. However, plastic pollution, which has been shown to have negative impacts on both human and environmental health, is becoming a global problem, with micro plastics being found in every corner of the Earth. Unfortunately, modern plastics are rarely recycled due to cost, degradation that occurs during mechanical recycling, and/or thermodynamic challenges present during chemical recycling. New "green" polymer designs that are non-toxic throughout the polymer life-cycle and that have suitable thermodynamic properties for cost effective recycling are needed to replace existing plastics. To find these designs, machine learning based property predictors will be combined with polymer design techniques to generate novel, hypothetical polymers that have the properties necessary for certain target applications while also being easily depolymerizable (an attribute that will allow the polymer to be recycled without property degradation). This work focuses on designing these plastics for food packaging and service items, which will require suitable mechanical, thermal, solubility and toxicity properties, to name a few. In addition (de)polymerization will be controlled by thermodynamics properties such as the enthalpy and entropy of (de)polymerization and ceiling temperature. Datasets for these properties will be curated and used to produce machine learning models for rapid prediction of the relevant properties. This predictive capability will be used to screen polymers that can undergo ring-opening polymerization (a promising polymerization mechanism for depolymerizable polymers) for those that meet targeted application needs, while also determining chemical guidelines for the design of these green polymers. Close collaborations with experimentalists to validate and realize several aspects of this effort are underway and will be strengthened further.