AI-Guided Inverse Design and Discovery of Recyclable Vitrimeric Polymers.

Area of Science:

• Materials Science
• Polymer Chemistry
• Computational Chemistry

Background:

• Vitrimers are sustainable polymers with dynamic covalent adaptive networks, offering self-healing properties.
• Limited molecular choices restrict vitrimer property tuning and application scope.
• Inverse design strategies are needed to tailor vitrimer chemistries for specific performance requirements.

Purpose of the Study:

• To develop an innovative computational framework for the inverse design of vitrimer chemistries.
• To enable precise control over the glass transition temperature (T_g) of vitrimers.
• To accelerate the discovery and synthesis of novel, sustainable vitrimers with desired properties.

Main Methods:

• Creation of a large-scale vitrimer dataset (one million chemistries).
• High-throughput molecular dynamics (MD) simulations to calculate T_g, calibrated by Gaussian process models.
• Development of a graph variational autoencoder (VAE) with dual graph encoders for multi-component vitrimer representation.

Main Results:

• The VAE framework accurately predicts T_g and discovers novel vitrimers with target properties beyond the training data.
• A novel vitrimer with a T_g of 311-317 K was synthesized and experimentally validated for healability and flowability.
• The computational framework demonstrates high accuracy and efficiency in designing vitrimers with desired glass transition temperatures.

Conclusions:

• The integrated MD simulation and VAE approach provides a powerful tool for designing custom vitrimer materials.
• This method facilitates the exploration of a vast chemical space for sustainable polymers.
• The framework empowers polymer chemists to synthesize novel vitrimers for diverse applications, enhancing material sustainability and performance.